video evidence on the effectiveness of patriot...

Science & Global Security, 1993, Volume 4, pp.I-63Photocopying permitted by license onlyReprints available directly from the publisher@ 1993 Gordon and Breach Science Publishers S.A.Printed in the United States of America

Video Evidence on theEffectiveness of Patriot duringthe 1991 Gulf War

George N, Lewiso and Theodore A, Postolb f,;Many of the Patriot-Scud engagements during the 1991 Gulf War were recorded on

video by news media camera crews. These videos are a unique and important source ofinformation on Patriot's performance that has been completely ignored by the USArmy in performing its assessments of Patriot's performance. The videos show manyexamples of Patriots missing their Scud targets, as well as a few exampl,es of Patriotshitting Scuds but failing to destroy the Scud warhead. The videos also contain infor-mation about the nature of the Scud targets, and also about other important occur-rences, such as Patriots diving into the ground. The videos we have been able to obtaincontain no unambiguous evidence that Patriot destroyed even one Scud warhead.Rather, they contain substantial evidence that Patriot's success rate was very low, andprovide a record of Patriot performance that clearly is inconsistent with the US Army'scurrent claim that Patriot destroyed 52 percent of the Scud warheads that it engaged.

INTRODUCTION

During the 1991 Gulf War, approximately 80 modified Scud missiles launched ,by Iraq landed in or near Israel and Saudi Arabia. The Army initially claimed lthat there were 47 instances of Patriots engaging Scuds; however, Congres-sional experts report that the current Army figure is slightly lower. Therewere 16 to 17 engagements in Israel and 27 to 29 in Saudi Arabia.!

During the war, and in the weeks that followed, the US Army maintainedthat Patriot had successfully intercepted almost all of the Scuds (45 out of47,or 96 percent). Since then, the Army has several times lowered its claims of

a. Defense and Arms Control Studies Program, Massachusetts Institute ofTechnology, Cambridge, Massachusetts.b. Defense and Arms Control Studies Program and Program in Sdence,Technology, and Society, Massachusetts Institute of Technology, Cambridge,Massachusetts.

2 Lewis and Postol

Patriot effectiveness, most recently to just below 60 percent overall-over 40percent in Israel and over 70 percent in Saudi Arabia. The Army also claimsPatriot destroyed 52 percent of the Scud warheads it engaged, but only 40 per-cent of the claimed warhead kills2 are in the "highest confidence" category.3Although the Army's data remain classified, they have been examined by sev-eral congressional agencies, which found the Army's data to be inadequate tosupport their claims of Patriot effectiveness.4

The Army's assessments completely neglect a crucial, and unclassified,source of information on Patriot performance: videos of Patriot-Scud engage-ments taken by neWs media camera crews.5 In this paper, we discuss what can

1 be learned about Patriot performance from news media videos. We will pro-ceed primarily by presenting a detailed analysis of videos showing the three

, engagements that occurred in Riyadh, Saudi Arabia on the nights of 25 and 26January 1991. We will then present the overall results of our studies of thenews media videos, which strongly suggest that the level of Patriot effective-ness during the Gulf War was very low.

NEWS MEDIA VIDEOS OF PATRIOT-SCUD ENCOUNTERS

The videos were taken by television camera crews located in the cities of Riy-adh and Tel Aviv, and near the large airfield in Dhahran, Saudi Arabia.6 Theengagements were frequently filmed from the hotels in which the cameracrews were staying.

Where possible, we have used videos which have not been edited for broad-cast; such unedited videos have been provided to us by ABC News and byWETA-TV in Washington DC.7 Other engagements were recorded directly offour televisions and have generally been edited for broadcast. In addition, aspart of an analysis conducted at the request of Congress, we have also ana-lyzed engagements contained in a collection of news broadcast videos that wasprovided to the Congress by the Raytheon Company.S In carrying out our anal-ysis, we have correlated the video evidence with both broadcast and printmedia press reports and with official statements by the US, Israeli, and Saudi

governments.The videos we have examined contain data that bear on slightly more than

half of the engagements. In some engagements, only part of the engagementhas been captured. An entire engagement might not be recorded because thecamera crew did not respond quickly enough to record the beginning of theengagement, or because the camera turned to another event before theengagement was completed, or because some of the engagement was obscured

.Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 3

by clouds. Appendix A contains a brief summary of many of the significantevents seen on the videos.

We have analyzed in detail 33 intercept attempts contained in 18 engage-ments. We define an intercept attempt as an attempt by one Patriot to destroya Scud; an engagement is the set of all intercept attempts against one particu-lar Scud. The engagements we have analyzed each contain between one andfour intercept attempts. The 33 intercept attempts we have analyzed consti-tute a substantial fraction {about 38 percent} of all the intercept attempts dur-ing the Gulf War.9

The news media videos contain a great deal of important, useful, andunique information on the performance of Patriot in the Gulf War. 10 The vid-

eos show many clear misses by Patriots as well as almost certain hits on Scuds{although without destroying the Scud warheads}. They also contain manyexamples of Scud warheads surviving engagements and exploding on theground, as well as Patriots diving into the ground and exploding. The videosalso contain substantial information on the Scud targets, for example, on theirbreakups during re-entry and on the dynamics of their re-entry-for example,some of the Scuds appear to have followed helical trajectories.

Limitations of News Media VideosThe videos considered here were taken by news media cameras, 11 not by high-

resolution cameras on test ranges, and accordingly have a number of limita-tions. However, as we demonstrate below, the videos do provide informationthat is of critical importance to an assessment of Patriot's Gulf War perfor-mance.

One limitation of the press videos is that generally there is little detailedinformation about the conditions under which the videos were taken, such asthe locations of the cameras, the directions they were pointing, or the focallengths being used. In addition, the cameras were not fixed in known orienta-tions, but were generally hand- or shoulder-carried and thus were continuallyin motion. Further, the videos also provide only a two-dimensional view ofthree-dimensional events.

The above limitations could be overcome to some extent in future studiesby using the records of the television networks and major video repositories toidentify the camera operators, who could then be interviewed to obtain addi-tional data, such as camera locations. The data obtained from such interviewscould be further enhanced by taking nighttime videos from the identified loca-tions with cameras of the same type used to record engagements. Such addi-tional videos could provide calibration information that could be helpful in

4 Lewis and Postol

interpreting the videos of Gulf War Patriot-Scud engagements. Coupled withinformation on Scud warhead impact locations-most of which remains classi-fied-such information on camera locations could establish the viewing geom-etry of the intercept attempts seen on the videos.

Another limitation is the relatively low frame rate of the press videos. UStelevisions show only 30 frames per second, and as a result there is an uncer-tainty in the relative positions of the fireball produced by the detonation of thePatriot warhead and the Scud. The hot gases produced by the detonation of (

the high explosives in the Patriot warhead are rapidly (in at most a few milli-seconds) stopped by air friction, traveling no more than roughly five meters.12The location of the center of the fireball therefore is uncertain by only aboutfive meters in the first video frame in which it appears and the fireball is effec-tively stationary in subsequent frames. Thus the center of the fireball pro-vides a fixed reference point for measurements. However, at a typicalintercept altitude of roughly 10-12 kilometers, the Scud is moving at about2.0-2.2 km sec-l, or about 70 meters per video frame.

The speed of the Scud produces an uncertainty in the location of the Scudat the time of the Patriot warhead detonation. This uncertginty arisesbecause the shutter on the video camera is open only for a small fraction ofeach 0.033 second video frame. Since the Patriot fireball is stationary, it isseen at the same position no matter when during the frame time the detona-tion occurred. However, the moving Scud is seen at the location it occupiedwhen the shutter was open. We refer to the resulting uncertainty as the "det-onation location uncertainty"; it is simply equal to the distance the Scudmoves between two successive video frames (roughly 70 meters). The camera,however, sees only the component of the Scud's motion that is in the plane per-pendicular to the line of sight of the camera. This component is given by (V /30) .sin (X = (70 meters) .sin (x, where V is the Scud's speed (in m sec-l) and (Xis the angle between the camera's line of sight and the Scud's velocity vector.

We will refer to the two-dimensional detonation location uncertainty seenon the television screen as the apparent detonation location uncertainty, andthe full three-dimensional distance as the actual detonation location uncer-tainty. We will use the same nomenclature to denote other distances, such asmiss distances, where a two-dimensional projection is seen on the television

i screen.A number of analysts have claimed that the limitations of the news media

videos, the detonation locatiGn uncertainty in particular, prevent the videosfrom being a useful source of information on Patriot performance. In Box Aonpage 11 we refute the argument that the detonation location uncertainty pre-vents determination of hits or misses. Other criticisms, such as that the vid-

Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 5

eos do not allow a precise measurement of miss distances or cannot determinewhere on a Scud's body a hit may have occurred,13 are simply irrelevant to anoverall assessment of Patriot's performance in the Gulf War. We will demon-strate that, despite their limitations, the news media videos can be used todetermine whether a Patriot hit a Scud, or whether a Scud warhead survivesan intercept attempt. The videos can show that some Patriot missiles weredirected to the wrong points in space, and that other Patriots dove into theground and exploded. Taken together, the videos are a unique and importantsource of data on Patriot's Gulf War performance, and they clearly and unam-biguously indicate that the US Army's current claims of Patriot effectivenessare incorrect.

ANALYZING THE VIDEOS

To illustrate how the videos were analyzed and the types of information thatcan be extracted from them, we will consider three engagements in detail.These engagements occurred in Riyadh on the nights of 25 and 26 January1991. These three engagements were selected because they contain examplesof the three main types of Patriot-Scud encounters seen on the videotapes:"clear misses" and two types of "fireball overlaps" (intercept attempts that aretoo close to classify as misses based only on the apparent locations of the Scudand Patriot at the time the Patriot detonates). The first type of fireball over-lap is one in which the Scud does not appear to be affected by the interceptattempt; the second is one in which the Patriot appears to hit and damage theScud but does not detonate its warhead.

The images appearing in this article were produced by digitizing signalsfrom a video cassette recorder.14 The data were then transferred from thevideo capture electronics to the memory and display of a Super- VGA graphicsadapter as 640 x 480 pixel and 256 color images. The images are constructedby interpolation of the 512 x 480 pixel, 24 bit per pixel captured raw videodata from the digitizing input electronics. This choice of image-capture set-tings allows the images to be printed with a ratio of width to height of four tothree, similar to that observed on a television screen, with pixels of equalwidth and height and with 256 colors or shades of gray. The images were thenprinted using a laser printer.

The images have been collected into sequences which we refer to as "VideoSequences"; these Video Sequences illustrate the nature of the informationthat can be obtained from the videos. Each image within the Video Sequencesis referred to as a "frame" and is labeled by a number and a letter. Thus

6 Lewis and Postol

r~ "frame 7 c" means the third image in Video Sequence 7.

Some adjustments were made to the image brightness, contrast, hue, andsaturation settings, which are standard controls on televisions, in order tomake certain details of events visible. For example, in the second of the threeengagements that are analyzed in this paper, four frames of the engagement(frames 5a, 5b, 5c, and 5d) are captured using extremely high-contrast andbrightness settings relative to those used in later frames in that engagement(frames 6a through 60). Such variations in settings make it possible to showboth the behavior of the faint arriving Scud target before it encounters theI

,,;" Patriot, and that of the much brighter exiting Scud target (and its subsequent

breakup) after it is hit by the Patriot. However, (with a few noted exceptions),

settings were not changed within sequences (except where they were adjustedto best show buildings on the ground), and there is no additional computer

, processing of the digitized video data other than the magnification of selected, images in Video Sequence 7.15

Riyadh, 26 January 1991The first engagement we consider occurred at about 10:50 PM (Saudi localtime) on 26 January, when a single Scud was engaged by two Patriots. Theengagement is shown on Video Sequences 1 through 4, which are taken froman uncut video clip extending from before the launch of the first Patriot untilafter the impact and explosion of the Scud warhead. 16

First Patriot MissThe video record of the engagement begins with the launch of a Patriot. 17 The

standard operating procedure for a Patriot battery was to fire two Patriots,typically spaced about three seconds apart, at each Scud. Frame 1a shows thefirst Patriot, its bright booster flame easily visible, rising above the Riyadhcity lights. The camera tracks the first Patriot and thus does not observe thelaunch or flight of the second Patriot. Twelve seconds after its launch, therocket motor on the first Patriot burns out, and it disappears from view.

In engagell.lents such as this one, where a complete flyout of a Patriot isseen (at least until its motor burns out), it is possible to combine the informa-tion from the video with simulations of Patriot interceptors and Scud targetsto produce a rough timeline and range versus altitude plot of the engagement.Such a plot necessarily will be very approximate since only a two-dimensionalview of the Patriot flyout is seen, but nonetheless can provide valuableinsights into what occurred during an engagement. In this engagement, such

Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 7~

Frame I a Frame I b t = -0.033

Framelc t=O.OOO Frameld t=O.IOO

Frame Ie t = 0.200 Frame If t = 0.300

Frame Ig t = 0.400 Frame Ih t = 0.500

Video Sequence 1: Riyadh. 26 January 1991. The first of two Patriots misses the Scud. Time (inseconds) is referenced relative to the detonation time of the Patriot.'f

I

8 Lewis and Postal

20,: 0,,.: 0.,: 0

15 :: Scud trajectory--o,, 0: Location of Scud -A Patriot warhead

v. : at first miss 0 ~ explodes 17 sec.~ ~ : .after launch.z E 10 Location of Scud.at~.7Ci ,g breakup and s~cond miss 0 ~ Patriot coast

~ ,, .: °b--- Patriot burnout,: 0 ~.

5 : 0 ~ Patriot powered flightq ~

0 [ ~ Patriot trajectory

: ~: ~:~:~

05 10 15

Rangekilometers

Figure 1: Estimated time line and range versus altitude plot for the engagement at Riyadh on26 January 1991. This plot is necessarily very approximate since only a two-dimensional viewof the Patriot ftyout is seen on the videos. This plot assumes that both Patriots flew the sametrajectory. There is one second between each symbol.

an analysis suggests that the first Patriot intercept attempt could have takenplace at an altitude of about 11.5 kilometers and the second one at about 9.5kilometers. Figure 1 shows the estimated timeline and range versus altitudeplot for this engagement. IS

'l\vo to three seconds after the Patriot burns out, the camera picks up theincoming Scud. At this point, the Scud is about 14 seconds from impact, andcan be estimated to be at an altitude of roughly 13 kilometers and to be mov-ing at roughly Mach 7 (about 2.0 to 2.2 km sec-I). The drag due to air frictionresults in the dissipation of many tens of megawatts of power, ten percent ofwhich is typically dissipated in the Scud's body, while the rest is carried off bythe wake. This means that megawatts of power are being delivered to the sur-face of the Scud, resulting in skin temperatures that are high enough to make

'lil~ Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 9

it easily seen by video cameras operating at visible wavelengths.One and a half seconds after the camera acquires the Scud, the warhead

on the first Patriot detonates. Frame lb shows the falling Scud one videoframe before the first Patriot warhead detonates. 19 Trailing behind the bright

Scud body is a luminous wake, probably composed primarily of incandescentdebris coming from the Scud missile body as it begins to break up due to aero-dynamic re-entry forces. The shape of the luminous wake makes it clear thatthe Scud has undergone or is undergoing a dramatic maneuver.

Frame lc is the next video frame, 0.033 seconds later, and it shows thefireball from the detonation of the Patriot warhead. The Patriot systemattempts to detonate a Patriot warhead close enough to the target missile'swarhead so that there is a high probability that the target warhead will bestruck by one or more fragments from the Patriot warhead, causing the targetwarhead to detonate. The distance from the target warhead at which Patriothas a high probability of striking the target warhead with a fragment is oftendescribed in terms of an effective kill radius. According the General Account-ing Office, in order to have a high probability of destroying a Scud's warhead,"the Patriot missile must detonate when it is within a few meters of theScud."ZO However, as we discuss in appendix C, the kill radius is stronglydependent on the intercept geometry. For example, since the Patriot warheadfragments are accelerated to at most 2.5 km sec-I, which is less than the clos-ing speed between the Scud and Patriot (typically 3.0 to 3.5 km sec-I), anexplosion behind the Scud warhead will not damage it.

The bright luminous fireball from the detonation of the Patriot warhead isclearly visible in frame lc, and appears on the video to be well behind theScud. It is important to note that the fireballs seen on the videos are not truefireballs-balls of hot radiating gas-but rather are complex mixtures of hotgas, smoke, and debris from the warhead and missile. As discussed in appen-dix B, the Patriot fireballs seen in the news media videos appear to be quitelarge, with diameters of order 100 meters. Thus the fireball radii seen on thevideos are much larger than the Patriot's shrapnel kill radius, so that thedetails of the Patriot's shrapnel pattern are not important in analyzing inter-cept attempts in which the Patriot's fireball does not overlap the Scud's posi-tion.

The next five frames, ld to lh, show every third video frame, thus there is0.1 seconds between these frames. These frames show the Scud falling awayfrom the Patriot fireball, apparently unaffected by the intercept attempt.

The analysis of this intercept attempt begins with the Patriot's fireball,which, as discussed earlier, is very rapidly stopped by air friction. Using thecenter of the Patriot's fireball as a stationary reference point, we can then

1 0 Lewis and Postol

measure (on the television screen) the apparent miss distance (the distancebetween the Scud and the center of the Patriot fireball in the first video framein which the fireball can be seen) as well as the distance that the Scud movesbetween successive video frames.21 The distance the Scud moves between twosuccessive frames is particularly significant, because, as discussed above, thisdistance is equal to the apparent detonation location uncertainty due to theframe time of the video.

Video Sequence 1 demonstrates, for the intercept attempt shown, that thedetonation location uncertainty does not prevent us from determining that thePatriot missed the Scud. We find that the apparent miss distance is aboutnine times the apparent detonation location uncertainty. Therefore, theuncertainty in the Scud's apparent location due to the low frame rate of thevideo camera is at most a small fraction of the apparent miss distance. Thusthe apparent miss shown in Video Sequence 1 is an actual miss. We thereforelabel this intercept attempt as a "clear miss." We encourage the reader to con-firm this conclusion by making measurements on Video Sequence 1.22

In all the intercept attempts we label as clear misses, the apparent missdistance is at least roughly three times greater than the detonation locationuncertainty, and in most cases is much larger (on average about ten times

larger).As discussed above, we cannot directly translate apparent miss distances

on a television screen into actual miss distances because the Scud motion perframe corresponds not to the actual Scud velocity V; but to V .sin a, and theangle a is not generally known. However, if, as Video Sequence 1 suggests,the Patriot detonated in or near the wake of the Scud (or where the Scudwould have been if it had not been moving on a helical trajectory),23 then thesame sin a scaling factor would apply to both the detonation location uncer-tainty and the miss distance. In this case, assuming a Scud speed of 70meters per frame (or about two km sec-I) at the intercept attempt, the actualmiss distance would be roughly 9 .70 = 630 meters. Such a very large misscould well be the result of the self-destruction of the Patriot after failing tofuze on the target.24

If the assumption that the Patriot detonated in or near the wake of theScud is not correct, then the miss distance must be scaled by the factor sin a.As discussed in appendix B, typical values for sin a are in the range 0.1-0.6(camera angle of 6° to 37°). The resulting actual miss distance is still manytimes larger than the kill radius of the Patriot warhead, and is far too large forthe Patriot to have had any effect on the Scud.

,

I!

I

Video Evidence on the Effectiveness of Patriot during the 1991 Gulf Waf 11

Box A: Criticism of the Use of Videos

A number of individuals and organizations have criticized the use of news media video-tapes in assessing the performance of Patriot.25 Their primary argument is that the deto-nation location uncertainty resulting from the low frame rate of news media videocameras makes it impossible to determine if a given Patriot actually hits or misses its Scudtarget. For example, the prime contractor for the Patriot program, the Raytheon Com-pany, has stated: "No analysis performed by or for Raytheon failed to support the com-pany's position that standard television video is an inadequate medium, because of itsslow frame rate, to determine any!hing definitive or scientific concerning the success orfailure of a Patriot-Scud intercept."26

Our argument is that, in all the cases we have labeled as clear misses, the apparentmiss distance is so large that the detonation location uncertainty is irrelevant. However,with one notable exception, the critics simply assert that because the detonation locationuncertainty exists, it is impossible to determine hits or misses; they do not attempt torelate the size of the uncertainty to the size of the misses seen on the videotapes. The oneexception is Peter Zimmerman, who uses an assumed Patriot fireball diameter of eightmeters to measure apparent miss distances on the videos. Zimmerman then compares theapparent miss distance with the detonation location uncertainty, which is about 70 meters.Zimmerman concludes:27

When the apparent miss distance as recorded on videotape is less than about fivefireball diameters (about 40 meters), it is impossible to make any valid judgmentabout whether the intercept was successful or not, based only on the separationbetween the fireball center and the Scud warhead. The event could have beeneither a hit or miss, and the video record provides no evidence either way.As the apparent miss distance increases towards ten fireball diameters, the prob-ability of Patriot's success slowly decreases. If the apparent miss distance exceedsten fireball diameters, the probability of the intercept being within the lethalrange of the warhead is sufficiently low that one may reasonably score theengagement as unsuccessful based only on the commercial videotape record.

Since few of the intercept attempts appear to be off by more than ten fireball diameters,Zimmerman's criterion implies that the news media videos of intercept attempts are of lit-tle utility in determining hits and misses. However, there are two serious flaws in Zimmer-man's argument. First, as is discussed in detail in appendix B, his assumed fireball diameteris far too small, probably by a factor of about 10. If he had used a more correct fireballsize, Zimmerman's approach would lead to the conclusion that the videos show many clearmisses.

Second, and even more important, Zimmerman compares the apparent miss distance(which will always be smaller than the actual miss distance) with the maximum possibledetonation location uncertainty of about 70 meters. Although he uses his fireball diame-ter length scale to measure apparent miss distances, inexplicably, he does not use thislength scale to measure the apparent detonation location uncertainty. The apparent deto-nation location uncertainty is smaller than the fireball diameter in all the engagements wehave categorized as clear misses (see table B-2). Thus if Zimmerman had used his eightmeter fireball diameter length scale to measure the apparent detonation location uncer-tainty, he would have found that the apparent detonation location uncertainty was alwaysless than eight meters and typically was two or three meters. Thus even using Zimmer-man's too-small eight meter fireball, the apparent misses seen on the videos are indeedactual misses.

12 Lewis and Postol

Scud BreakupVideo Sequence 2 shows the subsequent motion of the Scud at 0.1 secondintervals (except for 2h, which is only 0.067 seconds after 2g). The persistentand luminous Scud wake in the photographs strongly suggests that the Scudwas following a helical trajectory, with a period of about one second, as it fell.If so, then the Scud was subject to very high lateral accelerations as it fell,which would have greatly increased the difficulty of intercepting it.

The Scud's maneuvers most likely are a result of instabilities introducedinto the missiles when they were modified by the Iraqis to increase theirrange.28 The aerodynamic forces associated with these maneuvers as well asthe normal de-acceleration of the Scuds as they penetrate more deeply into theatmosphere, together with the poor quality of the Iraqi modifications, causedthe Scuds to breakup during re-entry, typically at altitudes at or aboveroughly ten kilometers.

The Scud breakup illustrated in Video Sequences 1 through 3 is a typicalone.29 The first clear sign of the breakup is the increase in brightness of theScud in frame 19. Variations in brightness continue in Video Sequence 2 andthe beginning of the formation of a luminous debris cloud associated with thebreakup is seen in frame 2e. In frames 2f to 2h, two objects can be seenemerging from the front of the luminous cloud of Scud debris. As we shall see,the object on the left is the dense Scud warhead section, which continues tofall at a high speed, pulling away from the breakup debris cloud, which is rap-idly stopped by the atmosphere. Frame 2h shows the situation one videoframe before the detonation of the second Patriot warhead.

Second Patriot MissVideo Sequence 3 shows the second Patriot intercept attempt, which occurred1.3 seconds after the first intercept attempt. Frame 3a is a repeat offrame 2h,showing the situation one frame before the detonation, and the detonation ofthe second Patriot warhead is shown in frame 3b. The fireball from thePatriot warhead detonation overlaps both the objects that were seen emergingfrom the debris cloud. We refer to such an intercept attempt in which theScud cannot be seen in the first frame following the Patriot detonation as a"fireball overlap."

The occurrence of a fireball overlap does not by itself establish whether ahit or miss occurred, because, as discussed in appendix B, the apparent size ofthe Patriot fireball is much larger than the kill radius of the Patriot warhead.

j Thus an overlapping fireball does not necessarily, or even probably, indicatethat the Patriot has hit the Scud. Following a fireball overlap, the behavior of

j


Frame 2a t = 0.600 Frame 2b =

Frame 2c t = 0.800 Frame 2d t = 0.900

Frame2e t=I.000 Frame2f t=I.100

Video Sequence 2: Riyadh. 26 January 1991. The Scud breaks up due to aerodynamic forcesand its warhead emerges from the breakup debris cloud.

14 Lewis and Postal

Frame 3a t = 1.267 Frame 3b t = 1.300

Frame 3c t = 1.400 rame t = .5

Frame 3e t = 1.600 rame t = .

Frame 3g t = 1.800 rame t = .

Video Sequence 3: Riyadh, 26 January 1991. The warhead on the second Patriot detonates,and the fireball overlaps the apparent position of the Scud warhead. However, the Scudwarhead continues onward.,~


the Scud target must be carefully observed for evidence of the effect of the

intercept attempt.Frames 3c to 3h show the events following the second Patriot warhead

detonation at 0.1 second intervals. These frames show the leftmost of the twoobjects that emerged from the Scud breakup continuing to fall, apparentlyunaffected by the intercept attempt. Eleven seconds after the second Patriotdetonation, this object-the Scud warhead-struck the ground, producing abright flash and a large fireball.3D The Scud warhead impact and explosion isshown in Video Sequence 4.

Frames 4a and 4b show the warhead falling to the ground. Frame 4c istaken one video frame before detonation; the warhead itself is lost among thelights on the ground. Light from the beginning of the Scud warhead explosioncan be seen in frame 4d. One video frame later, a large fireball as well as anarea of diffuse scattered light is apparent (frame 4e). By 0.5 seconds after theexplosion (frame 4g), the amount of scattered light has decreased sufficientlyto allow an apparently well defined "fireball" to become visible.31 This fireballis approximately 35 meters high and 100 meters across.32 The Scud fireballlooks much larger on the video than the Patriot fireballs both because theScud warhead contains several times more high explosive and because theScud detonation is much closer to the camera than the two Patriot detona-tions. Even two seconds after the explosion (frame 4h), some glowing rem-nants of this fireball are still visible (frame 4h).

The warhead struck in an area described by a television reporter as a"wasteland," and no casualties or damage were reported.33 The total time ofthe engagement, from the launch of the first Patriot until the impact and deto-nation of the Scud warhead, was about 31 seconds.

In summary, in this engagement two Patriots were fired at one Scud. Thefirst Patriot missed by a large distance. The second Patriot appeared toexplode relatively close to the Scud, and the apparent fireball from the detona-tion overlaps the Scud warhead position on the video. However, the Scud war-head was not destroyed and continued onwards, eventually exploding on theground. The US Army apparently categorizes this engagement as a success.34

Riyadh, 25 January 1991':--First EngagementThe other two engagements we discuss in detail occurred about 10-15 secondsapart on the night of 25 January 1991 in Riyadh. All publicly available infor-mation indicates that two Scuds were fired at Riyadh that night and that bothwere engaged by Patriot.

1

;:0 ~~ 16 Lewis and Postol

!J~

Frame 4a t = -0.433 Frame 4b t = -0.233

Frame 4c t = -0.033 Frame 4d t = 0.000

Frame 4e t = 0.033 Frame 4f t = 0.233

Frame 4g t = 0.500 Frame 4h t = 2.00

Video Sequence 4: Riyadh. 26 January 1991. The Scud warhead explodes on the ground.Time is referenced relative to the time of the Scud warhead explosion.


Camera ViewsWe have three different camera views of the first of these two engagements, inwhich a single Patriot attempts to intercept a Scud.35 We will use two of thethree different camera views in reconstructing the engagement.36

The first video is filmed from an unknown location and begins with thelaunch of a Patriot. The Patriot flies for about 9.3-9.4 seconds before itencounters a faint Scud target and detonates. The intercept attempt occurs ata relatively low altitude and the Patriot rocket motor is still burning when thePatriot warhead explodes.37

The second view of this intercept attempt was recorded by a cameralocated just outside of the Riyadh Marriott f{otel (an({, unfortunately, viewsthe intercept attempt over several street lamps). The video clip begins shortlybefore the detonation of the Patriot warhead, and thus does not include thelaunch and most of the flight of the Patriot.

Patriot DetonationVideo Sequence 5 shows the Patriot detonation and the three video framespreceding it, as seen by both cameras. Frames 5a to 5d are from the camera atthe unknown location, and frames 5e to 5h are from the camera at the Marri-ott. The Patriot appears to be moving nearly vertically towards the top of thetelevision screen, and the Scud appears to be moving downwards.

As seen by both cameras, the Patriot warhead appears to explode veryclose to the Scud's apparent position, and, as in the second intercept attempton the 26 January Scud, the Patriot fireball overlaps the Scud's position (andso we label this intercept attempt as a "fireball overlap"). In this case, how-ever, the behavior of the Scud target is dramatically and unambiguouslyaltered by the intercept attempt.

Video Sequence 6 shows the sequence of events as the Scud target emergesfrom the Patriot fireball, as seen by the camera at the unknown location.38Frame 6a is the video frame before the Patriot warhead detonates, however,due to the brightness and contrast settings used in Video Sequence 6, neitherthe Patriot nor the Scud can be seen.39 Frame 6b shows the Patriot warheaddetonation-this is the same as frame 5d except for the different brightnessand contrast settings. Frames 6c to 60 then show the next 13 video frames,each separated by 0.033 seconds.

By frame 6d, the Scud begins to become visible as a separate bright spotbelow the center of the Patriot fireball, although it was still overlapped by thefireball (see Video Sequence 7 for a more detailed look at these events). The

I

~~~ :l~;i,~;e 18 Lewis and Postol

:~

Frame 5a t = -0.100 Frame 5b t = -0.067

Frame 5c t = -0.033 Frame 5d t = 0.000

Frame 5e t = -0.100 Frame 5f t = -0.067

Frame 5g t = -0.033 Frame 5h t = 0.000

Video Sequence 5: Riyadh. 25 January 1991. first Scud. Two different cameras record aPatriot intercept attempt in which the Patriot fireball overlaps the Scud position. Time is refer-enced to the time of the Patriot warhead explosion.


Frame 6a t = -0.033 Frame 6b t = 0.000

Frame 6c t = 0.033 Frame 6d t = 0.067

Frame6e t=0.100 Frame6f 1=0.133

Frame 6g 1 = 0.167 Frame 6h 1 = 0.200

Video Sequence 6: Riyadh. 25 January 1991. first Scud. The Scud target emerges from thePatriot fireball with greatly increased brightness. then undergoes an abrupt increase in bright-ness, and then fades from view. Time is referenced to the time of Patriot warhead detonation.

20 Lewis and Postol

Frame 6i t = 0.233 Frame 6j t = 0.267

Frame 6k t = 0.300 Frame 61 t = 0.333

'-

rame m t = .Frame 6n t = 0.400

rame 0 t= .

Video Sequence 6.

"1


rame a t= .rame a agDll(.

Frame 7b t = 0.100 Frame 7b Magnified

Frame 7c t= .rame c agDlle

Video Sequence 7: Riyadh, 25 January 1991, first Scud. Selected images from VideoSequence 6 are magnified in order to more clearly show the Scud emerging from the Patriotfireball and then abruptly increasing in brightness.

22 Lewis and Postol

separation between the Scud and the center of the Patriot fireball increases asthe Scud continues onward, and by frame 6g the Scud is clearly visible as aseparate bright spot just below the Patriot fireball. It is evident that the Scudemerges from the intercept attempt with a greatly increased brightness.

In frame 6h, about 0.2 seconds after the Patriot detonation, the Scudabruptly and dramatically increases in brightness and it continues to growbrighter in frame 6i. The Scud then fades in brightness until, about a half-sec-ond after the intercept attempt, it is no longer visible. Moreover, after theScud flares up it also appears to stop moving relative to the Patriot fireball-the separation between the Scud and the Patriot fireball is essentiallyunchanged between frames 6h and 6n. This indicates that a Scud debris cloudformed and was rapidly stopped by air friction.

Video Sequence 7 shows magnified views of the Scud emerging from thePatriot fireball and flaring up. Frame 7a is the same as 6c-one frame afterthe Patriot detonation-and is shown magnified as frame 7 a(M). Frames7b(M), 7c(M), and 7d(M) are magnified versions of frames 6e, 6g, and 6i,respectively. The emergence of the Scud from the Patriot fireball and its sub-sequent flaring-up or explosion can be clearly seen in frames 7c(M) and 7d(M).

Interpretation of the VideoWe believe that the sequence of events described above provides strong evi-dence that the Patriot hit the Scud, although as we shall see, it apparently didnot destroy the Scud's warhead.

What would we expect to see if a Patriot hit a Scud and caused the Scudwarhead to detonate? The detonation of the Scud warhead would break thewarhead casing into many small fragments and would also heavily damageother parts of the Scud. Any remnants of the Scud's body would then break updue to the effects of aerodynamic forces. Such an event is unlikely to result ina single well-defined target emerging from the fireball, and an emerging objectwould never continue on a high-speed ballistic trajectory. In addition, the det-onation of the warhead and the disintegration of the Scud body would result inthe immediate development of a debris cloud. As with the Patriot fireball, thisdebris cloud would be rapidly stopped by atmospheric drag. If the Scud war-head detonated immediately, these events might be indistinguishable from thedetonation of the Patriot warhead and the formation of its fireball.4o Finally,there should be no ground explosion associated with the engagement.41

The intercept attempt shown in Video Sequence 6 has many of these char-acteristics, but is in other ways anomalous. Clearly the Scud warhead was notimmediately detonated by the Patriot, because the Scud continues onwards for

Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 23-

0.2 seconds, although the Scud's increased brightness indicates that it hadbeen hit and damaged. A delayed detonation of the Scud's warhead isunlikely, but is possible. The subsequent formation of the Scud fireball ordebris cloud as seen on the video also differs from what would be expected dueto the detonation of the Scud warhead, and in some ways more closely resem-bles a Scud breaking up. The Scud appears to increase in brightness over twoor possibly three video frames, whereas a fireball from warhead detonation istypically brightest in its first frame. Nevertheless, it is possible for a warheaddetonation to produce a pattern of brightness variation such as that demon-strated by the Scud in Video Sequence 6 if the timing of the camera shutterrelative to the warhead detonation is just right.42

There is, however, at least one other explanation for the events seen inVideo Sequence 6. An intercept attempt that does not damage a Scud's war-head could still lead to the Scud partially or completely breaking up. Anyintercept attempt that does serious damage to a Scud would alter its aerody-namic characteristics, producing a change in its aerodynamic heating rate andthereby in its brightness. Thus a hit that does serious damage to a Scudshould produce visible changes in the Scud's brightness and/or motion. AsVideo Sequences 1, 2, and 3 demonstrate, the breakup of Scuds due to re-entrystresses produces bright debris clouds43 often comparable in brightness to thefireballs produced by Patriot warhead detonations. Such a bright debris cloudcould occur if, for example, a Patriot detonated near a Scud, but its fragmentsstruck behind the Scud warhead. In this case, the explosion of the Patriotwarhead would tear the back part of the Scud body apart while leaving thewarhead intact. In fact, the Patriot detonation shown in frame 3b was misi-dentified as part of the Scud breakup in a report issued by the US Army'sWhite Sands Missile Test Range.44 The breakup of a Scud caused by a Patriotexplosion might produce even more spectacular visual results. Thus theevents of Video Sequences 6 and 7 are consistent with the Scud being damagedby the Patriot explosion, leading to its breakup 0.2 seconds later.

The Iraqi Scuds re-enter at higher speeds than the missiles Patriot wasdesigned to intercept. As we discuss in appendix C, there have been reportsthat Patriot was not fuzing properly against the Scuds. Appendix C alsoshows that based on the publicly available data on Patriot's warhead and fuz-ing system, it is plausible that Patriot's fuze could have been detonating thePatriot warhead too late to damage the Scud's warhead. If true, this would beconsistent with the second interpretation of this intercept attempt: the Patriotfuzed on the Scud but the warhead detonated too late and Patriot warheadfragments struck the Scud body, but hit behind the Scud warhead.

The anomaly associated with this explanation is that the warhead is not

~

,: :t\;j; 24 Lewis and Postol

Frame 8a t = 4.20 seconds Frame 8b t = 4.23 seconds

Frame 8c t = 4.33 seconds

Video Sequence 8: Riyadh. 25 January 1991, first Scud. The Scud warhead explodes on theground. 4.23 seconds after the Patriot intercept attempt.

seen continuing onward. However, there are several other instances on thevideos where warheads were not visible after normally occurring Scud break-ups or where warheads were not visible during Patriot intercept attempts.45

In the absence of further information, it might be difficult to choosebetween these two explanations of the events following the intercept attempt.As we shall see, however, the video record provides strong evidence that theScud's warhead survived the intercept attempt and detonated on the ground.

Frames 8a to 8c are a sequence of three video frames, respectively 4.20,4.23, and 4.33 seconds after the Patriot detonation. These frames are from thecamera at the Marriott.46 Despite the bright street lights and the camerabeing pointed above the horizon, a bright flash from a ground explosion isclearly seen in frame 8b-note the silhouette of the tree above the street signand the illumination of one side of the street lamp pole. All three camerasshow the ground flash occurring 127 video frames, or 4.23 seconds, after thePatriot warhead detonation.

Since, viewed from this location, the Scud appeared to have been fallingnearly vertically, this explosion appears to have occurred at the approximate


20,,,,......,,,

15 :II...II.I

QI ~: 0-QJ I

I

.-E E 10 : Scud trajectory ---0~ 0 .<~ : 0.,

.0,,

: 0, 0

5 : Patriot: trajectory 0

: 1 0: G-- Scud time to ground: '0 = 4.23 sec. \: A 60 Warhead impact point, 6~

05 15

Rangekilometers

Figure 2: Estimated timeline and range versus altitude plot for the first engagement at Riyadhon 25 January 1991. This plot is necessarily very approximate since only a two-dimensionalview of the Patriot flyout is seen on the videos. This plot indicates that a ground explosion 4.23seconds after the intercept attempt is consistent with the intercept attempt occurring about9.3-9.4 seconds after the Patriot launch.

azimuth where the Scud warhead would have impacted. The object in the skyat the top of frames 8a to 8c provides a reference point-it is a debris patternproduced by the Patriot's unburned solid rocket fuel.47

Moreover, the timing of the ground flash is also consistent with a detona-tion of the Scud's warhead. Figure 2 shows an estimated timeline and rangeversus altitude plot for this engagement. Assuming that the ground explosionwas due to the Scud, and that the Scud was on a trajectory such as thatdescribed in appendix B, then the intercept attempt would have occurred at analtitude of about 2.5 kilometers. The Patriot flew for about 9.3-9.4 secondsbefore its warhead detonated; because the Patriot launch site, the direction ofScud approach, and the first camera location (the camera that observed theentire flight of the Patriot) lie nearly in a line, it is not possible to determine

26 Lewis and Postol

the angle of the Patriot's trajectory. However, as figure 2 shows, a flight timeof about 9.3 seconds is entirely consistent with a Scud warhead explosion 4.23seconds after the intercept attempt. Thus it is highly likely that the observedground flash is due to the detonation of the impacting Scud's warhead. Aswith most of the Scud warheads, the impact point of the warhead is not pub-licly known and there were no reports of damage or casualties.48

In summary, in this engagement one Scud was engaged by one Patriot.The Patriot appears to detonate close to the Scud, and we label the interceptattempt as a "fireball overlap." The behavior of the Scud after the interceptattempt strongly suggests that the Patriot hit the Scud. However, the Scudwarhead reached the ground and exploded.

Other Video Evidence of Patriots Hitting Scuds

There is only one other intercept attempt (out of the 33 intercept attempts weanalyze) that shows the type of behavior seen in this instance-a Patriot fire-ball overlapping the Scud position, followed by a dramatic change in the Scudluminosity and with no target observed continuing on a high-speed ballistictrajectory. This other intercept attempt occurred at Riyadh on 11 February,when a single Scud was engaged by two Patriots. The first intercept attemptwas a fireball overlap, but there was no apparent effect on the Scud, whichcontinued onwards and then underwent an apparently normal breakup. Thesecond intercept attempt, however, produced a sequence of events very similarto those in Video Sequence 6, in that the Scud flared in brightness and thenbegan to fade. We have only a broadcast video clip of this engagement, andthe clip was cut shortly after the second intercept attempt so that we cannotbe certain whether or not the Scud warhead continued onwards. However, thecontinuation of the broadcast news story then cuts to an explosion on the

ground.There is no doubt that the warhead of this Scud hit the ground and

exploded. There was only one Scud fired at Riyadh on 11 February (and therewere none for several days before and after the 11th). The Scud warhead hitnear a school or university, producing a crater roughly three feet deep and 10-

-, 15 feet across and causing extensive damage to a nearby building, although

only two night watchmen suffered minor injuries.49 The video recorded at theimpact site includes scenes of the shattered concrete exterior walls of a steelreinforced building, clearly indicating an intense pressure wave due to an

explosion.In addition, two other intercept attempts captured on video may show a

Patriot hitting a Scud: the second intercept attempt on the infrared video

:ii

I1

t:


(date and place unknown) and the second intercept attempt at Tel Aviv on 19February. However, in both these cases the Scud warhead was seen to con-tinue onwards.

In the video recorded with an infrared camera, the Scud emerges from thefireball with a greatly changed appearance. It appears smaller, dimmer, andwith a much less pronounced wake. The emergence of the Scud from the fire-ball looks in some ways similar to what is seen at visual wavelengths when aScud warhead emerges from a normal Scud breakup. This suggests that thePatriot hit the Scud, although without destroying its warhead. However, thevideo clip ends shortly after the intercept attempt, so that the Scud warhead isnot tracked to the ground.

In the 19 February engagement, the Scud appears unaffected as itemerges from the fireball. However, about 0.6 seconds after the interceptattempt, the Scud breaks up, leaving a debris cloud, and the warhead contin-ues onward. This would be unremarkable, except that the video shows thatthe Scud had already broken up, raising the possibility that the Patriot wasresponsible for this unusual second breakup. None of the three videos record-ing this engagement show a ground explosion. According to Israeli militarysources, the Scud warhead did not explode, and there was no evidence of dam-age to the warhead due to Patriot fragments. Defective fuzes were believed tobe the most likely cause of the failure to explode.50

The US Army claims that Patriot caused three Scud warheads to fail toexplode. 51 However, despite recovering the warheads, the Army could not pro-

duce for Congressional investigators any chemical or metallurgical evidencethat Patriot was responsible for preventing the warheads from exploding, andin no case was a Patriot warhead fragment recovered from a dud Scud war-head.52 In addition, it is known that at least one Scud warhead that was notengaged by Patriot did not explode.53

Riyadh, 25 January-Second EngagementThe second engagement at Riyadh on the evening of 25 January occurredabout ten seconds after the first and is illustrated in Video Sequences 9, 10and 11.

Shortly after the ground flash from the detonation of the first Scud, thecamera at the Marriott suddenly swings upwards and to the left, where itacquires a second Scud about seven seconds after the intercept attempt on thefirst Scud. The camera tracks the Scud and, within a few tenths of seconds, aPatriot is seen to detonate, apparently behind the Scud. Frame 9a shows theScud one frame before the Patriot warhead detonation and frame 9b shows the

28 Lewis and Postol

Frame 9a t = -0.033 Frame 9b t = 0.000

Frame 9c t = 0.200 Frame 9d t = 0.400

Frame ge t = 0.600 Frame 9f t = 0.800

Video Sequence 9: Riyadh. 25 January 1991. second Scud. The first of two Patriots misses theScud


Frame lOa Frame lOb

Frame 10c Frame 10d

'# '#

I .J l

Frame 10e Frame lOr

Frame 109

Video Sequence 10: Riyadh. 25 January 1991. second Scud. The Scud is seen falling over theRiyadh Marriott hotel. chased by a Patriot. and a flash from an explosion is seen as the can-:-era turns away.

30 Lewis and Postol

Frame lla Frame lIb

Frame llc Frame lId

Frame lIe Frame llf

Frame llg Frame llh

Video Sequence 11: Riyadh. 25 January 1991. second Scud. The Scud is seen falling over abuilding, followed by a flash from a ground explosion. The explosion of the warhead of thechasing Patriot is seen 1.1 seconds after the ground explosion.


Patriot warhead detonation. The Scud is just visible in 9b at the edge of thePatriot fireball. This Patriot fireball appears unusually large and bright andsome of the apparent diameter of the fireball in 9b appears to be due to scat-tered light. By turning down the television brightness cqntrol, the size of thePatriot fireball can be reduced by a factor of about two relative to that shown

, in 9b, with the Scud still visible, now clearly separated from the fireball.54I Note that in this sequence the Scud is moving towards the top of the tele-, vision screen. This occurs because, as we shall see, the Scud flew nearly

directly over the camera. Frames 9c through 9f were taken 6, 12, 18, and 24video frames, respectively, after the Patriot detonation and show the Scud con-tinuing onwards in apparently normal fashion. In the first 24 frames after thedetonation, the total distance the Scud moves relative to the center of the fire-ball is only about half of the apparent fireball diameter in frame 9b (since theScud is on the edge of the fireball in 9b, half the apparent fireball diameter isalso equal to the apparent miss distance). As with the first intercept attemptanalyzed in this paper, this establishes that the ap'parent detonation locationuncertainty is only a very small fraction of the apparent miss distance.Accordingly, it is clear that this apparent miss is an actual miss and so welabel this intercept attempt as a clear miss. 55

Following the intercept attempt, the camera follows the Scud for abouttwo and a half seconds before the video clip abruptly ends. A second video clipapparently begins shortly after the end of the previous clip. As we shall see,this Scud passed almost directly over the camera, and we believe that the sec-ond clip is a continuation of the first clip after the camera operator has turnedto reacquire the Scud as it passes overhead. 56

Video Sequence 10 is taken from the second video clip. Frames lOa andlOb show the Scud flying over the Riyadh Marriott and disappearing behindthe building. Roughly one-half second after the Scud disappears behind thebuilding, the camera acquires a second object which appears to be followingthe Scud along the same general trajectory (frames 10c and 10d). As it moves,it does not slow up but becomes dim, following a brightness profile similar tothat of Patriots as their motors burn out. As the camera is turning away, 1.3seconds after the Scud disappears behind the Marriott, the video reveals aflash behind the Marriott (frame lOt).

Frame 109 shows how the location of the camera is known. After theevents described above, the camera pans around, revealing the entrance to theMarriott hotel.

Another video clip,57 filmed from a different location, shows the sameScud flying over and falling behind a large building (not the Marriott). This isshown in frames 11a and Ilb. About 1.4 seconds after the Scud vanishes

32 Lewis and Postal

behind the building, a bright flash from a ground explosion can be seen behindthe building (frame lId). A Patriot detonates in the sky above the building(frame lif) 1.1 seconds after this ground flash.

The Scud hit a six-story building in downtown Riyadh.58 The building,which belonged to the Saudi Interior Ministry, was demolished (frames ligand lih). Despite the building reportedly being unoccupied (the attackoccurred at about 10:20 PM on a Friday night), one person was killed andabout 30 injured. The building was about 2.25 kilometers south of the Marri-ott hotel. Since the Scuds were coming from the north, the Scud must havepassed almost directly over the camera crew at the Marriott at an altitude ofover two kilometers. The events associated with this Scud attack weredescribed by BBC reporter Barnaby Mason as follows, "Then another, appar-ently a Scud, came shooting right over our heads across the city and, in pur-suit of it, another missile-what appeared to be a Patriot-racing after it. ThePatriot apparently failed to catch up with the Scud which came down to theground with an enormous explosion.',59

In summary, the second of the two Scuds that arrived in Riyadh wasengaged by at least two Patriots. The first Patriot missed the Scud by a largedistance, and the second Patriot detonated in the air after the Scud hadalready hit the ground.

OTHER INFORMATION CONTAINED ON THE VIDEOS

The videos contain significant information on events other than Patriot hits ormisses. For example, as the 26 January engagement in Riyadh (in which theScud is seen falling with a helical motion) illustrates, the videos containpotentially very useful and important information on the Scud targets. Manyof the Scud breakups are recorded on the videos, and following some of thebreakups, showers of debris can be seen falling away from the breakup debriscloud. These breakups caused serious problems for Patriot by creating manypotential debris targets-yet it appears that this problem was not identifiedand corrected until after the attacks of 25 January, by which time more thanhalf of the engagements had already taken place.GO

One particularly important class of events captured on the video is Patri-ots diving into the ground and exploding. The videos we have been able toobtain show at least five Patriots diving into the groundGl (see appendix A)although we believe the total number of Patriots that dove into the ground islikely to be considerably higher.G2

Video Sequence 12 shows the first and second of a group of three Patriots


F,.,..I2a F~12b

Video Sequence 12: Tel Aviv. 25 January 1991. A Patriot dives into the ground and explodes inor near Tel Aviv. The Patriot explosion was followed by two additional explosions. the first ofwhich is shown here. The additional explosions may be secondary explosions caused by thePatriot explosion and/or Scuds exploding on the ground.

that was launched from the Patriot battery located at Tel Aviv's Dov Airport(just north of downtown Tel Aviv) at about 6 PM on 25 January 1991. (Thethird Patriot detonated prematurely a few seconds into flight.)

Frame 12a shows the bright explosion produced by the impact of firstPatriot on the ground. This Patriot flew a low arcing trajectory, striking theground about nine seconds after launch. The Patriot's contrail is illuminatedby the flash from the explosion, so that the Patriot's trajectory can be easilyseen. It seems likely that the trajectory that this Patriot planned to fly wastoo low and intersected the ground.63 After 25 January, the Patriot softwarewas changed to raise the minimum intercept altitude, and there are no reportsof Patriots diving into the ground after this date.

Immediately following the Patriot impact and explosion, two additionalbright explosions are seen. Frame 12b shows the first of these explosions,which occurred one second after the Patriot impact. These two explosions maybe secondary explosions produced by the Patriot impact or one of the explo-sions may be from the Scud that the Patriot was trying to intercept (and theother could be from a more distant Scud impact).

Video Sequence 13 (which is only one image) shows another example of thetype of information that can be obtained from the videos. Video Sequence 13shows the fourth clear miss on a Scud fired at Tel Aviv at about 7 PM on 11February 1991.64 The Patriot warhead detonation illuminates the Patriot'scontrail, which shows that the Patriot was attempting to turn and chase theScud warhead. This indicates that the Patriot system had initially sent thePatriot interceptor to the wrong point in space-strongly suggesting that seri-ous systematic problems remained with Patriot even quite late in the war.(Eighty-five percent of all the engagements had taken place by 11 February.)

J;;j;;

34 Lewis and Postal

Video Sequence 13: Tel Aviv, 11 February 1991. The fourth of four clear misses on a Scud isshown. The detonation of the Patriot warhead illuminates its contrail (as well as that of thethird Patriot), showing that the Patriot had attempted to turn and chase the Scud.

CONCLUSIONS

We have described three engagements in detail and briefly discussed severaladditional occurrences in order to illustrate the type of information that canbe obtained from a careful analysis of news media video tapes of Patriot-Scudengagements.

The news media videos are a substantial, and in many ways unique,source of hard physical data on Patriot performance in the Gulf War.Although much of the video data is fragmented, in many cases the videos pro-vide a fairly complete picture of individual engagements. They not only pro-vide substantial detail about the outcome of many engagements, but they alsocontain detailed information on the breakup and aerodynamic behavior of theScuds as well as information on related occurrences such as premature Patriotdetonations and Patriots diving into the ground. We believe that a compre-hensive search through the press video, particularly looking for uneditedvideo, would uncover a great deal more data.

With the videos we have been able to obtain, we have been able to classify33 individual intercepts as either "clear misses" or "fireball overlaps."65Twenty-five of the thirty-three intercept attempts were clear misses.

The other eight intercept attempts were "fireball overlaps." (Two of thefireball overlaps occurred on the same engagement.) In all of the fireball over-laps, either the Scud warhead is seen to continue onwards (four cases), or aground explosion is seen (two cases), or both (two cases). Thus, in all eight ofthe fireball overlaps, it is clear that the Scud warhead was not detonated inflight.

These 33 intercept attempts were drawn from 18 different engagements.

Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 35~ The videos show at least five other Scuds that were engaged hitting the

ground and exploding (see appendix A), and the videos show at least part ofI the flight of several other Scuds that were likely to have been engaged. In

addition, the videos also show at least five Patriots diving into the ground andexploding.

Overall, the videos we have analyzed contain significant information on atleast half of the Patriot-Scud engagements. (The exact number of engage-ments remains classified, but is no more than 46.) In all of this video, there isno unambiguous evidence of a Scud warhead being destroyed in the air. Aboutthe best claim that can be made for Patriot is that, in a few cases, the videodata cannot prove that Patriot did not cause the Scud warhead to become adud. However, no mechanism has been advanced that could produce such anoccurrence except as a relatively low probability event (particularly in caseswhere the Scud does not appear to have been damaged by the intercept

attempt).66Thus the videos contain no unambiguous evidence indicating that Patriot

destroyed even one Scud. The videos instead contain substantial evidencethat Patriot's success rate was very low, possibly even zero. Although the vid-eos cannot establish what Patriot's actual success rate was (if for no other rea-son than that they provide information on only about half the engagements),they do provide a record of Patriot performance that is inconsistent with theUS Army's current claim that Patriot destroyed 52 percent of the Scud war-heads that were engaged.

The information contained on the videos is of particular importance giventhat the Army's claims rely almost entirely on data that remains classified,and that the Army's data has been criticized by Congressional investigators asbeing incapable of supporting the Army's claims.67

One lesson relevant to assessing the future performance of Patriot hasapparently been learned: Patriots recently deployed to Kuwait reportedly areequipped with digital data recorders.68 The deployment of video cameras toroutinely record all future Patriot engagements should also be considered;such video cameras could very effectively complement the use of digital datarecorders in assessing the combat effectiveness of any future use of Patriot (atleast for those intercept attempts occurring at night), Had such video infor-mation been available and used during the Gulf War, some of the problemsPatriot had during the Gulf War might well have been identified earlier thanthey were.

The development and deployment of defenses against tactical and theaterballistic missiles seems certain to be a major element of defense policy in theUnited States and other countries in the near future. Tens of billions of dol-

36 Lewis and Postol

lars may be spent in this area in the next decade. It is essential that in thedevelopment of these anti-missile systems, the lessons of the only operationalexperience with missile defense, the use of Patriot in the Gulf War, be fullyunderstood and exploited. By continuing to deny the utility of the news mediavideos, the US Army is denying itself the benefits of a unique and importantsource of information on the performance of missile defenses and, even worse,may be proceeding with the development of future defenses based on a flawedassessment of Patriot's performance.

We believe that a thorough, independent review of Patriot's performancein the Gulf War is essential. This review should make use of all the availableinformation, both classified and unclassified, and should be conducted by areputable, independent organization, such as the National Academy of Sci-ences or the American Physical Society. Given the inadequate data collectedby the US Army, even such a review may be unable to yield a completely defin-itive verdict on Patriot's performance. But it will certainly provide a muchclearer picture of Patriot's performance in the Gulf War and the implicationsof this performance for future efforts to defend against ballistic missiles.

ACKNOWLEDGEMENTS

We would like to thank the Ploughshares Foundation for very rapidly provid-ing us with funds for our research on Patriot at two critical points in our work.Ploughshares' contributions were absolutely critical to the success of ourresearch program. We would also like to thank the John D. and Catherine T.MacArthur Foundation, the John Merck Fund, and the W. Alton Jones Foun-dation for funding various parts of the work which led to this paper. Duringpart of the period during which this work was performed, one of us {Lewis}was supported by a Research and Writing Grant from the MacArthur Founda-tion's Program on Peace and International Cooperation. Dr. Richard L. Gar-win's continual support, encouragement, and critical review of our work hasbeen invaluable throughout the process of our research. Drs. Dan Fensterma-cher, Lisbeth Gronlund, and David C. Wright provided us with numerousvaluable technical and editorial comments. Finally, we would like to thankDr. Harvey Sapolsky, Director of the M.I. T. Defense and Arms Control StudiesProgram, for his unwavering support, encouragement, and constant goodhumor throughout what was often a very trying process.

NOTES AND REFERENCES

1. See John Conyers, Jr., "The Patriot Myth: Caveat Emptor," Arms Control Today,November 1992, pp. 3-10, note 3 and Eric Schmidt, "Israel Plays Down Effectiveness ofPatriot Missile," New York 1}:mes (International Edition), 19 October 1992, p. A8;


Ethan Bronner and John Aloysius Farrell, "US, Israeli Experts Dispute PatriotClaims," Boston Globe, 19 March 1992, p. 1.

2. If we assume a total of 46 engagements, then the Army's claimed success rateslead to the following:

Israel 17 engagements. 40% = 7 successful engagementsSaudi Arabia 29 engagements. 70% = 20 successful engagementstotal = 27 successful engagements

Since the Army claims that 52% of the engagements were warhead kills:46 engagements. 52% = 24 warhead kills

The above calculations lead to the conclusion that in three engagements claimedas successes, Patriots diverted Scuds ofT their trajectory and out of the defended areawithout destroying the Scud warheads or caused the Scud warhead to detonate withless than full yield (such a result was termed a "mission kill").

3. However, the Army's classification of an engagement as a high confidence war-head kill is misleading. The assignment of a high confidence warhead kill to a particu-lar engagement only means that the Army has higher confidence of a warhead killrelative to other engagements in which the Army believes there is some evidence ofsuccess. The GAO review of the Army's claims reports that, "According to the DeputyProject Manager, the assignment of a high confidence level to an engagement's outcomedid not mean that the Army was absolutely confident that the assessed outcome wascorrect. Rather, given the limited data available for assessment purposes, the Armyscorers had higher confidence in the assessed outcome of these engagements than inothers." US General Accounting Office, Operation Desert Storm: Data Does Not Exist toConclusively Say How Well Patriot Performed, GAO/NSIAD-92-340, September 1992,p.3.

4. US General Accounting Office, Operation Desert Storm: Project Manager's Assess-ment of Patriot Missile's Overall Performance is Not Supported, GAOtr-NSIAD-92-27, 7April 1992; Steven A. Hildreth (Congressional Research Service), Evaluation of USArmy Assessment of Patriot Antitactical Missile Effectiveness in the War Against Iraq,Testimony prepared for the House Government Operations Subcommittee on Legisla-tion and National Security, 7 April 1992; and US General Accounting Office, OperationDesert Storm: Data Does Not Exist. All of the above reports are reprinted in Hearingbefore the Legislation and National Security Subcommittee of the House GovernmentOperations Committee, Performance of the Patriot Missile in the Gulf War (WashingtonD.C.: US Government Printing Office, 1993). See also Conyers, "'!'he Patriot Myth."

5. A US Army report concluded that the news media videos were not useful in deter-mining whether a Patriot hit or missed a Scud. Army Material Test and EvaluationDirectorate, White Sands Missile Range, Analysis of Video Tapes to Assess PatriotEffectiveness (Rev 1), 31 March 1992.

6. There are also a small number of videos from Haifa, Israel. However, we haveseen no videos from the other major Scud target, the large military complex at KingKhalid Military City in Saudi Arabia.

7. An entire engagement, from the launch of the first Patriot interceptor to theimpact of the Scud warhead on the ground, typically lasts about 25 seconds. Broadcastvideo is generally cut to show only what the news media judges to be the most interest-ing parts of the engagement, such as Patriot launches, the detonation of Patriot war-heads, or Scud explosions on the ground. Unedited video typically allows a much morecomplete reconstruction of events. The videotapes provided to us by ABC and WETA-TV contained both edited and unedited footage.

38 Lewis and Postol

8. George N. Lewis and Theodore A. Postol, An Evaluation of the Army Report" Anal-ysis of Video Tapes to Assess Patriot Effectiveness," Dated 31 March 1992: A Study Per-formed in Response to a Request by Congressman John Conyers, Jr:, Chairman of theHouse Government Operations Committee, (Cambridge, Massachusetts: M.I.T. Defenseand Arms Control Studies Program, September 1992).

9. Although 158 Patriots were fired during the Gulf War, approximately 24 werelaunched accidentally ("were launched at empty airspace") while about another 47were fired at Scud debris. Thus only about 87 Patriots were actually fired at Scud tar-gets. Joseph Lovece, "Electronic Noise From US Gear Prompted Errant Patriots,"Defense Week, 28 September 1992, p. 1.

10. The US Army did not use video cameras to record any of the Patriot engagements.Further, digital data recorders were used in only three of the roughly 47 engagements.The Israelis did, however, set up high-resolution test-range video cameras near TelAviv and recorded a number of engagements there (although many of the engagementsnear Tel Aviv took place in cloudy conditions).

11. One of the videos (containing two intercept attempts on one Scud) is almost cer-tainly not from the news media: it appears to have been taken with an infrared cam-era. At present we do not know the source of this video, although it is known that theIsraelis did record some engagements using infrared cameras. The infrared video wasbroadcast on the program "Admiral William J. Crowe: The Lessons of Modern War,"broadcast on the Boston PBS station WGBX at 11 PM on 10 January 1993.

12. Lewis and Postol, "An Evaluation of the Army Report," pp. 63-65.

13. For example, such arguments against the use of the news media video tapes havebeen made by analysts at the White Sands Missile Range (where miss distances can bedetermined to within 0.3 meters). Army Material Test and Evaluation Directorate,"Analysis of Video Tapes."

14. Each video frame is captured as a digital picture that is 512 pixels wide and 480pixels high. The color information associated with each digitized pixel is stored as a 24bit number, which can potentially represent 16.7 million colors or shades of gray.Although only eight bit (256 color) data was used for the pictures in this article, the full24 bit raw data from the video board is readily captured and stored as digitized imageswith our current system.

15. Thus frames 109, Ilg, and Ilh use different settings than the other frames of theVideo Sequences in which they are contained. In addition, frames 5a through 5d usedifferent settings than 5e through 5h (which were taken by a different camera).

16. This video clip was provided to us by WETA-TV in Washington DC.

17. For background information on the Patriot system and the Scud targets, see The-odore A. Postol, "Lessons of the Gulf War Experience with Patriot," International Secu-rity.163 (Winter 1991/92), pp. 119-171.

18; The assumptions used in calculating the Scud trajectories used in this paper aredescribed in appendix B. The Patriot trajectory shown in figure 1 is calculated byassuming that the Patriot first flies up nearly vertically and then attempts to fly up theScud's expected trajectory.

19. It appears that the Patriot detonation actually began just as the video camera'sshutter was closing for this frame. This is indicated by a diffuse patch of light along-side the Scud's wake; this patch is in the same location as the Patriot fireball in the


next frame.

20. US General Accounting Office, Operation Desert Storm: Data Does Not Exist, p. 6.The GAO also reported (p. 7) that, "[The Patriot Project Office] Chief Engineer saidthat Patriot's fuze can sense its target and detonate at up to six times the requiredmiss distance, resulting in an extremely low or no probability of kill. However, the sys-tem would still record a kill."

21. Because the available data does not allow a precise determination of the fireballdiameter (and because it is also likely that the observed fireball diameters may varydepending on factors such as intercept altitude and camera settings), we do not use thefireball diameter as a length scale in determining hits and misses or in measuring missdistances. In addition, in some of the brighter Patriot fireballs observed in the videos,where there is a considerable amount of light scattering around the fireball, the appar-ent size of the fireballs are strongly dependent on the brightness and contrast settingsof the television (especially in the first few frames after the warhead detonation). Seeframes 5d and 6b of this paper, which show the same video frame with different bright-ness and contrast settings, and between which there is a factor of two or more differ-ence in the fireball diameters. (However, the Patriot fireballs in the engagement weare now discussing do not appear unusually bright and their size is not dependent onthe television settings.)

22. A simple way to roughly analyze this engagement is to assume that the Scud's tra-jectory is a straight line passing through the center of the fireball (this assumption willlead to only a small error in this case). For example, first measure the apparent missdistance (the distance between the center of the fireball and the Scud) on frame lc-about 6 millimeters. Then measure this same distance in frame If (nine frameslater}--about 12 millimeters. Thus the Scud moved about 6 millimeters in nineframes. Thus the video uncertainty is about 0.7 millimeters and the miss distance isabout 6/0.7 z 9 times the video uncertainty.

23. In cases where we observe clear misses, it is apparent, given the large size of missdistances, that the Patriots could not have fuzed on the actual Scud targets. Indeed, itseems likely that many of the clear misses we see involve the Patriot self-destructingafter failing to fuze on the Scud target. Since the Patriot generally attempts to inter-cept the Scud head-on (that is, the Patriot is flying the Scud's trajectory in reverse),this self-destruct would be expected to occur in or near the Scud's wake. Alternatively,some of the clear misses might involve the interception of debris trailing behind theScud, in which case the Patriot detonation would also be expected to occur in or nearthe Scud's wake.

24. When an engagement failed, the Patriot missile was commanded to self-destructafter a pre-set time delay. US General Accounting Office, Operation Desert Storm:Data Does Not Exist, p. 6.

25. Army Material Test and Evaluation Directorate, Analysis of Video Tapes; Peter D.Zimmerman, Report for the House Government Operations Legislation and NationalSecurity Subcommittee on "Patriot Effectiveness (Rev ])" and Other Related SubjectsConcerning Patriot ATBM Performance During Operation Desert Storm, 14 September1992. See also affidavits by Lawrence S. Silverman and Kerns H. Powers, which arereprinted in House Government Operations Committee, Performance of the PatriotMissile, pp. 328-341.

26. Letter from James W. Carter (Vice President, Raytheon Company) to US Repre-sentative John Conyers, Jr., Chairman of the House Committee on Government Opera-

40 Lewis and Postol

tions, 14 August 1992, p. 11. This letter is reprinted in House Government OperationsCommittee, Performance of the Patriot Missile, pp. 309-327.

27. Peter Zimmerman, Report for the House, p. 13.

28. Postol, "Lessons of the Gulf," pp. 126-130.

29. As far as can be determined from publicly available information, most or all of theScuds fired by Iraq during the Gulf War, whether engaged by Patriot or not, broke upon re-entry. The persistent and incandescent wake trailing behind many of the Scudsstrongly suggests that debris was coming off the Scuds throughout much of the timethey were in the atmosphere. Thus it may be more accurate to think of the Scuds asundergoing a series of breakups, with the breakup such as the one illustrated in VideoSequence 2 being the "main" breakup (the one in which the warhead section separatesfrom the Scud missile body).

30. The ground explosion and fireball were also observed by another camera at a dif-ferent location. However, cases such as this where the camera is relatively close to theimpact point and has a clear line of sight to the impact point are relatively rare. Theonly other such case we have seen is the Scud that landed on or near the airbase atDhahran on 23 January (there are also two additional video clips in which theDhahran 23 January explosion is seen as a flash occurring behind nearby buildings).In most cases, the explosion of the Scud warhead is seen only as a flash on or over thehorizon.

31. Although it may appear that a well-defined fireball is visible in frames 4e and 4f,when the video is viewed on a color television, it is apparent that much of the apparentfireball diameter in these frames is actually due to scattered light.

As we will see below, this fireball is far too large to be a true fireball (a ball of hotradiating gas). The exact nature of what is seen here (and when a Patriot warheaddetonates) is unclear, but undoubtedly involves complex physical phenomena.

32. These dimensions can be established by noting that the width of the fireball in 4gis somewhat greater than the distance the Scud moves in the six video frames between4a and 4b (this must be measured with respect to stationary ground features, becausethe camera is moving). The speed of the Scud at impact is not known precisely, but afigure of 700 m sec-l is reasonable (this is the impact speed for the trajectory shown onfigure 1 and discussed in appendix B-an intact Scud would hit the ground at about 1.5km sec-l). The warhead will hit the ground at an angle from the vertical of about 48°,so sin <X is at least 0.74. Thus the camera sees the Scud moving at 0.70. 0.74 = 0.52 kmsec-l = 17 meters per frame, for a total of roughly 100 meters between 4a and 4b.

33. Due primarily to the inaccuracy of the Scuds, only a relatively small fraction ofthe Scuds (less than one in four) are reported to have caused casualties or significantground damage. Despite this, the video record contains a number of cases where aflash from a ground explosion is seen and for which it is known that the Scud that pro-duced this flash caused casualties or significant damage. In Tel Aviv, such cases aretwo Scuds on 25 January, a Scud on 9 February, and a Scud on 12 February. In Riyadh,these cases include the second Scud on 25 January, and single Scuds on 3 and 11 Feb-ruary. In Dhahran, the explosion of the Scud that destroyed the US military barrackson 25 February is also on the videos. These cases provide clear examples of what dis-tant Scud warhead explosions look like.

34. Former US Representative Frank Horton (who was the ranking minority memberon the House Government Operations Committee during that Committee's investiga-tion into the performance of Patriot) recently stated, apparently based on the classified

I

i~


us Army assessment, that this engagement was successful. Frank Horton, "ThePatriot Debate: Part 2," Arms Control Today, January/February 1993, pp. 26-29.

35. All three of the camera views of this intercept attempt are from the tape providedto us by ABC.

36. The third view of this intercept attempt is very similar to the second viewdescribed below, and possibly was filmed from the same location.

37. The Patriot's rocket motor burns for about 12 seconds. The only other interceptattempt we have seen where the Patriot's motor was still burning (or had just burntout) when the Patriot warhead exploded was the fourth intercept attempt on the Scudat Tel Aviv on 11 February (see Video Sequence 13).

38. Both of the other cameras show an identical sequence of events.

39. Frame 6a is the same as frame 5c, but with different brightness and contrast set-tings.

40. It might be expected that the fireball produced by the detonation of a Scud war-head would be considerably larger than that produced by a Patriot warhead detona-tion, since the Scud warhead contains several times more high explosive (roughly 200kilograms) than does the Patriot (roughly 40 kilograms). However, given the widevariation in apparent Patriot fireball diameters seen on the videos, it is not clear thatthis difference would be discernible.

41. The argument has also been raised that a Scud could be hit by one or a few Patriotwarhead fragments that do not detonate the Scud's warhead but result in the Scud'swarhead not detonating when it strikes the ground. However, to our knowledge, noone has proposed a mechanism which could produce such a result except as a relativelylow probability and thus infrequent event.

42. However, in such cases there is normally a very small bright core to the fireball inthe first frame that is surrounded by an area of scattered light. The flaring of the Scudhere does not have that appearance.

43. Such debris clouds are readily visible in the press videos. They are typically pro-duced during a time interval of tenths of seconds as a Scud breaks up, and once createdthey remain visible in the press videos for seconds.

44. This engagement is catalogued as tape VMS-8 at 33:14 in Army Material Test andEvaluation Directorate, Analysis of Video Tapes, Appendix A.

45. In addition, this intercept attempt occurs at a low altitude (less than fuve kilome-ters). Thus the main Scud breakup, which typically occurs at or above an altitude often kilometers, should have already occurred. However, little is known about whathappens in such breakups and it is possible that there is enough of the Scud remainingwith the warhead section to produce a second large breakup debris cloud.

46. The shoulder-carried camera has been moving towards the street light seen inVideo Sequence 8 (the same one seen in frames 5e to 5h), and the large round lightseen at the bottom of frames 5e to 5h is now just behind the camera.

47. Such a long-lasting debris pattern is not seen following Patriot detonations thatoccur after the Patriot's solid rocket motor has burned out (which is what normallyoccurs). However, the debris pattern is similar in appearance to that seen when Patri-ots detonate prematurely after only a few seconds of flight.

48. The videos we have analyzed show eight ground explosions in or near Riyadh-

.

I

42 Lewis and Postol

seven due to Scuds and one due to a Patriot that dove into the ground. Three of theseven observed Scud detonations produced reports of casualties and damage to build-ings, three did not, and one probably did not. The only other reported instance of casu-alties and damage in Riyadh probably resulted from a Patriot that dove into theground.

49. This description is based on reporting from the impact scene and video of theimpact scene and damaged building that was broadcast on CNN and ABC.

50. Private communication from Reuven Pedatzur.

51. Letter from Major General Jay M. Gamer, US Army, to Representative JohnConyers, Jr., Chairman of the House Government Operations Committee, p. 10. Thisletter is reprinted in Government Operations Committee, Performance of the PatriotMissile, pp. 277-308.

52. According to the Chairman of the House Government Operations Committee, "Inthe Army assessment, a dud Scud scored as a warhead kill if a Patriot had attemptedan intercept. However, many of the Scuds were duds to begin with. Scuds were foundwith concrete warheads, or little explosive, or broken wires in the fuzing section. Sev-eral of these were scored as kills, even without corroborating evidence such as radardata. The duds were often burned and broken from impact, but this was hardly "clearphysical evidence of Patriot intercept damage," although in one case an Army officerthought a Patriot fragment caused a hole. This opinion was not supported by anychemical or metallurgical analysis or recovery of a fragment. Duds not engaged byPatriot showed similar damage." John Conyers, Jr., "The Patriot Debate: Part 2," ArmsControl Today, January/February 1993, pp. 27,29.

53. On 19 January, before Patriot was operational in Israel, a Scud warhead struck amulti-story building in downtown Tel Aviv. The warhead was recovered intact from aground floor jewelry store. See George N. Lewis, Steve Fetter, and Lisbeth Gronlund,Casualties and Damage from Scud Attacks in the 1991 Gulf War, Working Paper 93-2,Defense and Arms Control Studies Program, M.I. T., March 1993, p. 29.

54. It is unclear why some fireballs appear to be unusually large and bright on thevideos. Brighter fireballs might be due to detonations that occurred relatively close tothe camera or to variations in video camera settings.

55. There are at least two reasons for the very large ratio of fireball diameter toapparent Scud motion per frame in this intercept attempt. First, as discussed above,the apparent fireball diameter as seen on the video may be larger (by a factor of abouttwo) than the "true" fireball because, in the frames shortly after the detonation, thereis a great deal of light scattering in the air around the fireball which makes the actualextent of the fireball difficult to measure. Second, this Scud landed relatively close tothe camera-about 2.25 kilometers away. Thus when the Scud is first acquired by thecamera, its angular position will change relatively slowly.

56. The two video clips follow immediately one after the other on the videotape pro-vided to us by ABC. However, it is also possible that the second clip was taken by thethird camera. Following the intercept attempt on the first Scud, the clip from the thirdcamera is cut. It is immediately followed by two short video clips, the first one showinga Scud rapidly streaking across the sky, and the second one is a repeat of the video clipfrom which Video Sequence 10 was taken. In either case, as we will see, the camerawas at the Marriott.

57. This clip was broadcast on NBC Nightly News, 25 January 1991.


58. Malcolm W. Browne, "2 Office Buildings Leveled in Riyadh; Body Recovered," NewYork nmes (International Edition), 26 January, p. 16; Richard Owen and ChristopherWalker, "'IWo Killed and 70 Hurt in New Scud Onslaught," London nmes, 26 January1991, p. Ie. (One of the deaths and many of the injuries cited in the title of the Londonnmes article occurred in attacks on Israel on the same night.)

59. Owen and Walker, "'IWo Killed and 70 Hurt."

60. The commander of the US Patriot forces in Israel testified that: "We deployed withsoftware version 33, which we had confidence in, but a lot of unknowns about becauseit was brand new to us.

"Now, within three days [the first Patriot engagement in Israel was on 22 January-three days later is 25 January] we found out it was flawed. It was flawed not simplybecause it wasn't engineered well. It was engineered against a different kind of threat.And if I can put it into context, in the United States, we were trained and the systemwas designed against a system that pitched a fast ball. What we got was a system thatthrew a knuckle ball. It was not a simple process to understand exactly what was hap-pening.

"But we within a few days-the soldiers who represented the United States ofAmerica and Israel made changes to that system based on their own ingenuity thatcaused us to stop the problems we experienced on 25 January." Testimony of ColonelDavid Heebner, House Government Operations Committee, Performance of the PatriotMissile, p. 234.

Starting on 26 January, the Patriot batteries in Israel switched from operating inautomatic mode to operating in semiautomatic mode. This change allowed Patriot toattempt to intercept only the fastest falling object to emerge from the Scud breakup-the Scud warhead section.

61. 'IWo Patriots are seen on the videos diving into the ground in Saudi Arabia andthree in Israel. There are several other instances on the videos of Patriots diving intothe ground, however, we believe that these are likely to be different camera views ofthe five Patriots cited above.

62. Representative John Conyers, Jr., Chairman of the House Government Opera-tions Committee, submitted the following written question to the US Army: "There arereports of eight PATRIOTs hitting the ground in Israel alone, are these reports cor-rect?" The Army's response is classified. (House Government Operations Committee,Performance of the Patriot Missile, p. 279) In private communications with three Israe-lis, we have been given the figures of 8, 9, and 11 for the number of Patriots that doveinto the ground in Israel.

63. The Patriot that dove into the ground in Haifa followed a similar, although per-haps somewhat higher, trajectory. However the other three Patriots seen diving intothe ground followed significantly different trajectories. Each of these three Patriotswas heading upwards until it abruptly made a sharp turn and dove into the ground.This behavior may reflect some other problem with the Patriot system.

64. All four intercept attempts on this Scud were clear misses. It is not known whyfour Patriots, all from the same battery, were fired at this Scud.

65. We believe that all of the 33 intercept attempts are unique events, but we cannotabsolutely rule out that two of them (one clear miss and one fireball overlap), containedin two engagements, could be different views of other intercept attempts already eval-uated. (In appendix A, these two engagements are listed as the third Scud at Dhahranon 20/21 January and the first "date-and-place-unknown" event).

44 Lewis and Postol

66. There were three engagements containing fireball overlaps in which no groundexplosion was seen on the videos. In one of these cases (in appendix A, this is listed asthe third Scud at Riyadh on 21 January), the Scud did not appear to be affected by theintercept attempt, but the camera did not track the Scud to the ground. In anothercase, (the third Scud at Dhahran, 20/21 January) the Scud again did not appear to beaffected by the explosion, but was tracked most or all of the way to the ground and noexplosion was seen. However, this Scud appeared to have landed at a large distancefrom the camera, and it is possible the flash of a ground explosion was not seen for thisreason. (Another Scud that appeared to land quite far from the camera (Tel Aviv, 12January) had a ground flash that was only faintly visible). Moreover, some Scuds thatwere not engaged by Patriot are also known to be duds. The third Scud was the one toTel Aviv on 19 February; it is discussed in the text. See also the discussion in note 52.

67. See the sources cited in note 4.

68. "Raytheon Wins Kuwaiti Contract," Aviation Week and Space Technology, 18 Jan-uary 1993, p. 21.

Appendix A: Video Tape Summary

This appendix summarizes many of the significant events seen on the videos we haveanalyzed. Of these, 33 intercept attempts that were categorized either as "clearmisses" (25) or "fireball overlaps" (8) are listed as such.

Tel Aviv25 January At least 2 Scuds are seen hitting the ground and exploding. In addi-

tion, 2 Patriots are seen hitting the ground in Tel Aviv (and also 1 inHaifa).

26 January 2 Patriots fly up into the clouds, then 1 Scud comes down out of theclouds and explodes on the ground.

9 February 3 clear misses, and the Scud warhead explodes on the ground.

11 February 4 clear misses, but the camera does not track the Scud to the ground.

12 February 1 clear miss, and the Scud warhead explodes on the ground.

19 February 1 clear miss and 1 fireball overlap, with the second Patriot possiblyhitting the Scud. The Scud continues onward, and the cameraappears to track it to the ground, but no ground explosion is seen.

Riyadh21 January First Scud Only the end of the Scud's flight is seen, and its war-

head explodes on the ground.

Second Scud 1 Patriot explodes behind clouds, the other Patriotdives into the ground and explodes, and the Scud war-head falls and explodes on the ground.

Third Scud 1 fireball overlap and 1 clear miss. The Scud continues

J ) .


onward but the camera does not track it to the ground

22 January Four Patriots and probably two Scuds seen. It is difficult to interpretwhat is occurring; the Patriots do not seem to be engaging the Scudsthat are seen.

25 January First Scud 1 fireball overlap. The Scud was almost certainly hit,but the Scud warhead explodes on the ground.

Second Scud 2 clear misses, and the Scud warhead explodes on theground.

26 January 1 clear miss and 1 fireball overlap. The Scud warhead continuesonward and explodes on the ground.

3 February 2 clear misses, the Scud warhead continues onward and explodes onthe ground.

S February 2 clear misses but the camera does not follow the Scud to ground.

11 February 2 fireball overlaps, and the second Patriot almost certainly hits theScud, but the Scud warhead explodes on the ground.

24 February First Scud 2 Patriot detonations, but the Scud is not visible untilwell after the intercept attempts.

Second Scud 1 premature Patriot detonation (a few seconds intoflight) and 1 clear miss, but the camera does not followthe Scud to the ground.

Dhahran

20/21 January (it is not known if the "third" Scud occurred before or after the othertwo):First Scud Is only seen briefly towards end of its flight but is not

followed to the ground.

Second Scud 1 clear miss, camera does not follow the Scud toground.

Third Scud (this may be a view of one of the other two Scuds from amuch more distant camera, but this seems unlikely)-1 fireball overlap. The Scud continues onwards, andthe camera tracks it most or all of the way to theground, but no ground explosion is seen.

A Patriot is also seen diving into the ground and exploding.

23 January 1 premature Patriot detonation, 1 clear miss, and 1 Patriot detonationbehind clouds. The Scud warhead falls to the ground and explodes.

26 January 3 clear misses, and the Scud warhead explodes on the ground.

25 February An unengaged Scud explodes on the ground. This is the Scud that hitthe US barracks in Dhahran.

,i

i

t:J',;"

"

~~: \§j 46 Lewis and Postol"'c.! Date and Place Unknown: 1 clear miss, then the Scud disappears (po

cloud), followed by a Patriot detonation when Scud is not visible. Theclip ends immediately after the second Patriot detonation.

Date and Place Unknown: 1 clear miss and 1 fireball overlap. The Scud warhead con-tinues onward, but clip ends soon after second intercept attempt.This is infrared video and its source is not known. The second Patriotmay have hit the Scud without detonating the Scud's warhead.

Appendix B: Intercept Geometry and Fireball Size

In this appendix, we discuss what we can learn about the apparent size of Patriot fire-balls from careful measurements and analyses of the news media videos. The Patriotfireball diameter is of interest for two reasons. First, the erroneous claim that thePatriot fireball is as small as eight meters has played a central role in the argumentsmade by critics of using the news media videos as a source of data about Patriot's per-formance. Second, a relatively large fireball size would indicate that the shrapnel pat-tern of the Patriot warhead should not playa central role in analyzing the videos.However, we stress that we do not, and never have, used the fireball diameter as a dis-tance scale for determining hits and misses.

As noted in the main text, the fireballs seen on the videos are not simply balls ofhot gas, but rather are complex mixtures of combustion products, smoke, and warheadand missile debris. Test-range videos of Patriot intercept tests, taken with precisioncameras during daytime lighting conditions, typically show Patriot fireballs with diam-eters of about 20-30 meters. Figure B-1 is a photograph of such a test-range inter-cept,I showing a fireball with a 25-30 meter diameter.2 However, the availableevidence indicates that the nighttime fireballs recorded by the news media cameras inIsrael and Saudi Arabia appear to be at least several times larger than the daytimetest-range fireballs.

As discussed in the main text, the video cameras see only a component of theScud's velocity, given by V .sin a, where V is the Scud's speed and a is the anglebetween the Scud's velocity vector and the line of sight of the camera. Ifwe knew bothV and a, then we would have a length scale with which to measure the fireball diame-ter, since the motion of the Scud between two consecutive video frames would corre-spond to a distance of (V. sin a)/30.

The velocity of a Scud can be at least roughly estimated by modeling its trajectory.Given such a model of the Scud's flight path, the angle a could then be estimated if weknow the location of the camera, the Scud's impact point, and the time between theintercept attempt and the Scud's impact on the ground.

Many of the Scud impact points are known in Israel; however, most of the inter-cept attempts caught on video occurred in Saudi Arabia, where very few impact pointsare known. However, there are a few cases where enough information is available toallow some conclusions to be drawn regarding the size of the Patriot fireballs.


Figure 8-1: A Patriot intercept attempt on a Lance missile at the White Sands Missile Range.The Patriot has damaged the control surfaces of the Lance (which will later cause it to tum-ble). but otherwise the Lance is intact. A measurement of the length to diameter ratio of theLance (a Lance is about 6.14 meters long and 0.56 meters in diameter) indicates that theLance is not significantly foreshortened by viewing angle effects (alternatively. the fireballdiameter can just be measured with respect to the Lance diameter). The fireball diameter isfour to five times the Lance length. or about 25-30 meters.

12 February, Tel AvivWe first consider the Patriot-Scud engagement at Tel Aviv on the morning of 12 Febru-ary. In this engagement, a single Patriot appears to miss a Scud by a large distance,with the Scud continuing onward and exploding on the ground. The location of theScud impact is approximately known.3 It fell between two houses in a neighborhood ofindividual homes in or near the town of Savyon. It leveled both houses and irreparablydamaged a number of other homes.

The location of the video camera is not precisely known, but it was certainly on anupper floor of a tall building in downtown Tel Aviv, most likely at the Hilton hotel.Savyon is about ten kilometers from downtown Tel Aviv, about 200 south of due eastfrom the major hotels located along the Mediterranean shore. The Scuds approachedfrom about 100 north of due east. For simplicity, we will assume that the Scud trajec-tory (and thus the intercept point), the impact point, and the camera location all lie in

r

;~~~~~i ;:~;;'f: 48 Lewis and Postol

if~"'!,

25

20

15QI ~ ~ = 3,000 pst---o

"C~.-2 E Scud trajectory--o

~ ~ Postulated Scud breakup {)10 Patriot trajectory ---6

""

13 = 1,000 pst --65 ""

00 5 10 15 20 25

Distancekilometers

Figure B-2: Estimated video camera viewing geometry for the engagement at Tel Aviv, 12February 1991. This figure assumes thot the Scud's direction of approach, the intercept point,and the camera location all lie in a single plane. The figure also assumes that the Scudbreaks up at an altitude of 11.4 kilometers, reducing its ballistic coefficient ~ from 3,OCX:J to1 ,OCX:J pounds per square foot. The camera sees a component of the Scud's velocity Vat thetime of the intercept which is given by V. sin a.

a single plane. This assumption reduces the intercept geometry to a two dimensionalproblem, and will lead to a slight underestimate of the fireball diameter.

The geometry of this intercept attempt is shown in figure B-2. The choice of theScud trajectory is important, because the fireball diameter will depend on the interceptaltitude and on the Scud's speed at the time of the intercept attempt. The Scud trajec-tory shown on figure B-2 assumes that an intact Scud, with a ballistic coefficient of3,000 PSF (pounds per square foot), breaks up at an altitude of about 11.4 kilometers.The warhead section is then assumed to continue onwards, but with its ballistic coeffi-cient reduced to 1,000 PSF. The resulting trajectory is in good agreement with what isknown about the time it took for Scud warheads to reach the ground following inter-cept attempts.4 Table B-1 provides relevant numerical data for this Scud trajectory.

In the 12 February engagement, the Scud is intercepted at an unusually low alti-tude, and the Scud warhead strikes the ground only 5.6 seconds after the intercept

~

-Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 49

Table B-1: This table lists the characteristics of the Scud trajectory used in this paper.The Scud is assumed to have a ballistic coefficient of 3,000 PSF (pounds per squarefoot) until it descends to an altitude of about 11.4 kilometers. At this point, the Scudis assumed to break up, reducing its ballistic coefficient to 1,000 PSF.

Time to Altitude Downrange Re-entry Speedimpact distance angle

km sec-l kilometers kilometers degrees km sec-l

13.0 11.38 11.36 43.6 2.04

12.0 10.02 9.94 43.8 1.90

11.0 8.76 8.62 44.0 1.75

10.0 7.59 7.42 44.3 1.61

9.0 6.51 6.32 44.5 1.48

8.0 5.52 5.31 44.8 1.35

7.0 4.61 4.40 45.1 1.24

6.0 3.77 3.57 45.5 1.13

5.0 2.99 2.81 45.8 1.03

4.0 2.28 2.12 46.2 0.95

3.0 1.62 1.50 46.6 0.87

2.0 1.01 0.93 47.1 0.80

1.0 0.45 0.41 47.6 0.74

0.0 0.00 0.00 48.0 0.69

attempt. For the Scud trajectory shown in figure B-2, the intercept attempt wouldoccur at an altitude of about 3.5 kilometers and at a ground range from the camera ofabout 13.3 kilometers. Thus the angle between the line of sight from the camera to theScud and the horizon is about 14.7°. At the intercept attempt, the Scud's velocity vec-tor is oriented about 45.6° above the horizon, so IX is about 30.9°. The Scud's speed isabout 1.09 km sec-l, so the apparent distance the Scud moves per frame is about(1,090/30) .sin (30.9°) = 18.7 meters. In this case, the fireball diameter is about sixtimes the Scud motion per frame (see table B-2). We thus get a fireball diameter of

about 110 meters.5In this engagement, the apparent miss distance is roughly 19 times the distance

the Scud moves per frame, so that the minimum possible miss distance is about 350meters. It is possible that the actual miss distance is much larger, since we see only a

two dimensional projection.

25 January, RiyadhNext consider the last of the three engagements discussed in detail in the main text-

I:

50 Lewis and Postol

Table B-2: The ratio of the Patriot fireball diameter to the apparent distance the

Scud moves between two consecutive video frames.o The intercept attempts

shown here are all of the ones we label as ~clear misses. (except for the one

recorded by the infrared camera) where this ratio could be directly measured.

Date Place Intercept attempt Ratio

9 February Tel Aviv 1 3,4

2 1.93 3.1

11 February Tel Aviv 1 2.4

2 1.5

4 1.5

12 February Tel Aviv 5.9

19 February Tel Aviv 11.5+

21 January Riyadh, Scud 13 6.9

25 January Riyadh, Scud 12 1 15.4

26 January Riyadh 3.2

3 February Riyadh 1 5.4

2 3.1

8 February Riyadh 1 2.4

2 1.2

24 February Riyadh 4.2

20/21 January 1.4

26 January Dhahran 1 5.0

2 3.33 1.6

unknown unknown 1 17.9

Average 4.8

Average ratio if the three intercept attempts with ratios 3.2

greater than 10 are dropped

a. In order to be conservatNe in our analysis. we have used (with one exception) the smoHest firebaH size seen on the firstfive video frames offer the Patriot warhead detonation (t,""s is why. tor example, the ratio for the second Scud to Riy-adh on 25 January is only about 15. instead ot the value ot 24 c~ed in the main text-which was obtained using thePatriot firebaH diameter in the first trame offer the detonation) Since the firebaHs are generally somewhat asymmetric.the diameters gNen here are the aveJage of the fireball's "length" and "width," The exception nated above is onecase where the cameJO just catches the beginning ot the warhead detonation. which is visible as a smaH bright point.For this case. the smallest fireboH diameter on frames 2 through 5 offer the detonation is used


20 ,,,,,

Scud trajectory --015

~ = 3,000 pst--o,: 0

~ : Postulated Scud -A

~ ~ : breakup V~ (lJ ,.~ E 10 : 6~ a '"=: '

~ : 6,: 6-- Patriot trajectory,: 6,

5 : 6: 6--~ = 1,000 pst: 6: 6

66 : Location ot video camera

6 ,6 '0

5 0 5 10 15

Distancekilometers

Figure B-3: Video camera viewing geometry for the second engagement in Riyadh on 25January. The figure assumes that the Scud impact point. the intercept point. and the videocamera location all lie in single plane. No intercept point is shown because the intercept alti-tude is not known.

the second Scud at Riyadh on 25 January 1991. The camera viewing the first interceptattempt on this Scud (shown on Video Sequences 9, 10, and 11) was located just outsidethe Riyadh Marriott Hotel. The Scud struck a Saudi Ministry of Interior buildinglocated about 2.25 kilometers south of the Marriott. Since the Scuds were coming fromthe north, this Scud must have passed nearly directly over the camera at an altitude ofabout two kilometers.

Figure B-3 shows the geometry of the intercept attempt, assuming the same Scudtrajectory used in figure B-2. The only problem here is that we cannot directly esti-mate the intercept altitude, since there is a cut in the tape between the interceptattempt and the ground explosion.

Nevertheless, we can proceed by calculating the angle a and the fireball diameteras a function of intercept altitude. Proceeding in the same way as in the previousexample, we get the plot of fireball diameter versus altitude shown on figure B-4. Thisshows a minimum fireball diameter of roughly 200 meters, assuming a relatively highintercept altitude of about 12 kilometers.

,

.., 52 Lewis and Pastal -

500

400

~ 300OJ

EIV ~.-(l)00'--

(l)

~ E.cf! 200~

100

00 2 4 6 8 10 12

Intercept altitudekilometers

Figure 8-4: Estimated Patriot fireball diameter as a function of the altitude of the interceptattempt for the first intercept attempt on the second Scud at Riyadh on 25 January 1991. Thefireball diameter is calculated using the geometry shown in figure B-3 and is calculated as afunction of intercept altitude since the intercept altitude is not known.

The analysis leading to figure B-4 also indicates that the minimum possible missdistance is about 200 meters. But the minimum possible miss distance is so large inthis case that the Patriot could not possibly have fuzed on the Scud. This strongly sug-gests that the Patriot explosion we see was due to the Patriot self-destructing afterhaving failed to acquire its target (or possibly due to the Patriot intercepting debrisfrom the Scud's breakup), which indicates that the Patriot detonation probablyoccurred in or near the Scud's wake. In this case, most or all of the sin (X factor wouldbe cancelled out, leading to an actual miss distance many times greater than 200meters.

The two cases analyzed here can provide only very rough estimates of Patriot fire-ball diameters, in part due to the erratic trajectories followed by some of the Scuds cap-tured on video. Nevertheless, the cases analyzed here clearly indicate that theobserved nighttime Patriot fireball diameters are at least several times larger than theroughly 25 meter diameter Patriot fireballs observed in daylight conditions.

It is also informative to consider what a review of the broader body of data on


~Azimuth of

arriving Scud

Location of Location ofvideo camera Scud impact

Figure 8-5: Geometry used in estimating the value of {X (the angle between the Scud's veloc-ity vector and the line of sight from the camera to the Scud) as a function of the position ofthe Scud's impact point relative to the camera. The model assumes that intercepts occur atan altitude of ten kilometers and that the Scud falls at a constant angle of 45°. It is thenstraightforward to calculate {X as a function of the distance of the Scud impact point from thecamera and the angle III between the azimuth of the direction of approach of the Scud anda line from the camera location to the Scud impact point.

intercept attempts indicates about the Patriot fireball diameter. Table B-2 shows theratio of the Patriot fireball diameter to the Scud motion per frame for 21 interceptattempts. The intercept attempts in table B-2 are all of the ones that we label as clearmisses where this ratio could be directly measured.6 Table B-2 shows that in everycase, the Patriot fireball diameter is larger than the distance the Scud moves perframe. Given that these intercept attempts are from 14 different engagements, with avariety of different viewing geometries, this strongly suggests that typical Patriot fire-ball diameter is comparable to or somewhat larger than the distance a Scud moves inone video frame time (or about 70 meters at a typical intercept altitude often kilome-ters).

Table B-2 shows that the ratio of the Patriot fireball diameter to the apparentScud motion per frame ranges from 1.2 to about 18. On average, the Patriot fireball isabout 4.9 times larger than the apparent distance the Scud moves per frame. If we

54 Lewis and Postol

exclude the three intercept attempts where the ratio of fireball diameter to Scudmotion per frame is very large (greater than 10) the average value of this ratio is about3.2.

In order to interpret these ratios, we need to know the range of typical values forthe angle IX. This can be estimated by constructing a simple geometrical model. Thismodel relies on two assumptions:

.The Scud falls from the intercept point to the ground at a constant angle of 45°. Asthe Scud trajectory on figures B-2 and B-3 suggest, this assumption is likely to bereasonable for many of the engagements that are on the video tapes.

.The intercept attempt takes place at an altitude of ten kilometers. This altitudeis a reasonable choice for a typical intercept attempt, although it is likely thatsome intercept attempts are a few kilometers higher, and it is known that somewere lower.

The geometry used in our model is shown in figure B-5. The video camera islocated at the origin of an x, y, z coordinate system (the z axis is vertical). The Scud isassumed to be approaching in an x-z plane (that is, its ground track is parallel to the yaxis). The Scud's impact point is defined in spherical coordinates by its distance R (inkilometers) from the camera and by an angle <I> measured from the x-axis. Thus <I> = 0corresponds to a Scud that is directly on line with the camera but falls short, whereas 11>= 1800 corresponds to a Scud that flies directly over the camera.

The coordinates of the impact point (in kilometers) are then given by x = R .cos 11>,y = R .sin 11>, z = 0 and the intercept point is simply at 45° up from the impact point: x =R .cos <I> + 10, y = R .sin 11>, z = 10. Since the Scud's velocity vector is assumed to be 45°from the horizontal, it is then a simple matter to compute the angle IX. Figure B-6shows the result: it plots the value of sin IX as a function of the angle 11> to the impactpoint for several different values of R (the distance of the Scud impact point from thecamera).7

How far from the camera did Scuds typically land? Consider the case of Tel Aviv.The exact or approximate impact locations are known for 16 Scuds that landed in thegeneral vicinity of Tel Aviv.8 Most of camera crews were apparently filming from thehotel they were staying at, and for the purpose of discussion here, we will assume thatthe camera was located at the Tel Aviv Hilton. The points corresponding to 14 of theseScuds are plotted on figure B-6.9 As can be seen on figure B-6, the values of sin IX rangefrom about 0.14 to 0.57, with an average of about 0.30.10

The product of the average ratio of fireball diameter to Scud motion per frame andthe average value of sin IX can be thought of as giving an "average" fireball diameter interms of the distance the Scud moves in one video frame time (roughly 70 meters at atypical intercept altitude often kilometers). This product is (3.2 to 4.9) .0.30 = 0.96 to1.47, indicating a fireball diameter of order (0.96 to 1.44).70 = 67 to 103 meters.


1.0

0.9

-a Ge~metrical f?ctorsc 0.8 Distance 0 for Impact points In

~ (in kilometers) Tel AVIV, Israel

0 0.7 of Scud i.mpact~ from video~ camera""iij 0.6v

":+"QIE 0.50QIC\ro 0.4...QI

E~ 0.30QI

"C::> 0.2 0

0.1

0.00 30 60 90 120 150 180 .

Angle between the direction of a Scud impact pointfrom an observing video camera and the azimuth of

the arriving Scud

degrees

Figure 8-6: The value of the geometrical factor sin IX as a function of the distance of the Scudimpact point from the video camera and the angle II> between lines drawn from the camerato the Scud impact point and the Scud's azimuth of approach. The geometry used to calcu-late IX is shown in figure B-5. The four solid lines are for Scuds assumed to land respectively 2.5.5, 10, and 20 kilometers from the camera. Only positive values of II> are shown, since sin IX issymmetrical about II> = O. The circles show the values of 11>, sin IX, and the distance of theimpact point from the camera for 14 Scud impact points in or near Tel Aviv. assuming that thecamera is located at the Tel Aviv Hilton.

56 Lewis and Postol

NOTES AND REFERENCES1. This photograph is identified as: "Patriot Intercepts Lance Missile," 11 September1986. US Army Photo by: Optics Land/Air, WSMR, NM, photo # 6." This is thus a PAC-1 warhead intended for use against airplanes; however, we are aware of no reason whythe PAC-2 warhead used against the Scuds should produce a significantly different sizefireball.

2. In figure B-1, the fireball is four to five times the length of the Lance {which is6.14 meters long-indicating a fireball with a diameter of 25-30 meters. The length todiameter ratio of the Lance in figure B-1 establishes that the Lance is nearly intactand is viewed from near side-on.

We have obtained test-range video showing eight Patriot intercepts of missiles(including, it appears, the one shown in figure B-1). These videos are not as clear asthe original of the photograph reproduced in figure B-1, but rough measurements offireball diameters are possible. In two cases, the fireball diameter-missile length ratiowas about four, in four cases it was about five, in one case about 10, and in the othercase more than 20 (the reason for this very large fireball is not known-it may be dueto unburned fuel in one of the missiles or to some similar cause). No corrections havebeen made for angular factors-the videos generally appear to be side-on views-andin at least some cases, the relevant missile length is that of a Patriot (5.3 meters)rather than that of a Lance, and it cannot be conclusively ruled out that one or two ofthe videos are different camera views of the same intercept attempt.

The test-range intercept attempts are from the following three videotapes: "PatriotATM Capability Deployed for Multi Threat Capability," Missile Systems Division, Ray-theon Company, Bedford, Massachusetts; "After Desert Storm," Independent Commu-nication Associates, Ltd., London; and "The Patriot Missile: Hero or Hoax," an episodeof the Arts and Entertainment Network's "Investigative Reports" series.

3. Lewis, Fetter, and Gronlund, Casualties and Damage.

4. Typical intercept altitudes were apparently about ten kilometers. The averagetime to ground for Scuds following an intercept attempt was about 11.8 seconds. Thetrajectory assumed here has a time to ground of 12 seconds from a ten kilometer alti-tude.

The average time to ground for Scuds following an intercept attempt was deter-mined by averaging over 12 of the 14 intercept attempts for which this time could bemeasured. Two intercept attempts that clearly took place at unusually low altitudeswere excluded from this average. These are the first intercept attempt in Riyadh on 25January, and the intercept attempt at Tel Aviv on 12 February-the one we are dis-cussing here. The twelve times to ground used were (in seconds):

Riyadh 26 January 12.4, 11.1Dhahran 26 January 13.3, 11.4, 9.3Tel Aviv 9 February 15.5, 12.3, 12.0Riyadh 3 February 12.8, 11.5Dhahran 23 January 11.1,8.9

5. Ifwe had assumed that the Scud behaved like an intact Scud (with a ballistic coef-ficient of 3,000 PSF), then the intercept altitude would have been 6.7 kilometers, theScud speed at the intercept attempt wo\tld have been 1,890 m sec-l, and (X would havebeen 22.4°, giving a fireball diameter of about 140 meters.

6. In a few of the clear misses, the Patriot fireball and the Scud cannot both be seenon the television screen at the same time. In addition, the clear miss on the infrared


video is not included, because the appearance of the fireball is significantly differentfrom that of the fireballs recorded by cameras operating at visual wavelengths.

7. Since sin (X is symmetric for positive and negative values of <1>, we plot sin (X onlyfor the positive values of <1>.

8. A map showing the impact locations plotted in figure B-6 is in Lewis, Fetter, andGronlund, Casualties and Damage, p. 23.

9. Two Scuds that landed at larger distances (15 and 16 kilometers) from the Hiltonare not plotted because it is not clear if Patriot would have attempted to engage them.Note that not all of the Scuds plotted on figure B-6 were engaged: some of them fellbefore Patriot was operational.

10. The points plotted in figure B-6 are primarily at angles between 0 and 90°. Largerangles are uncommon because the camera location is near the coast; most Scuds withhigher angles would have fallen into the Mediterranean. Since higher angles tend togive higher values of sin (x, it seems likely that a somewhat higher average value of sin(X would be obtained if it were possible to include Scuds that fell into the Mediterra-nean.

Appendix C: Patriot Warhead Kill Radius and the Patriot Fuze

This appendix discusses some of the fuzing-related factors that determine whether it ispossible for a Patriot to successfully destroy the warhead of an incoming Scud missile.Since the end of the Gulf War, there have been credible and persistent reports thatPatriot was not fuzing properly against Scud targets during the Gulf War.! In thisappendix, we will show that publicly reported technical details about the properties ofthe Patriot's fuze and its warhead indicate that it is entirely plausible that thesereports are correct and that the Patriot fuze may have been detonating the Patriotwarheads too late to destroy the Scud warheads.

The Patriot PAC-2 warhead and fuze were developed for use against missiles withranges comparable to that of the Scud-B, which has a range of about 300 kilometers,and they were tested against such targets. However, during the Gulf War, Patriot wasattempting to engage 600 kilometer range modified Iraqi Scuds, which re-enter at aspeed that is approximately 40 percent higher than the shorter-range Scud-B. As aresult, the closing speeds between the Patriots and the Iraqi Scuds were higher thanthe speed for which the Patriot fuzing system had been designed. Because of this highclosing speed, it is possible that the Patriot fuze may have been detonating the Patriotwarhead too late to destroy the Scud warhead.

The Patriot system is designed to destroy its target by detonating a fragmentationwarhead close to the target. When a Patriot warhead detonates at the proper time,steel fragments from the Patriot warhead are sprayed out towards the target causing itto be damaged or destroyed. As we will show below, the fragments from a Patriot war-head travel at a speed that is lower than the closing speed between the Scud andPatriot. As a result, unless the Patriot warhead is detonated with near perfect timing,the front end of the Scud (which contains the warhead) will be missed by the Patriotfragments and the Scud warhead will be undamaged. We will show below that thePatriot PAC-2 fuze cannot sense an arriving Scud until it is nearly at the last possible

Ii

!

i.

!i

58 Lewis and Postol

point where a Patriot warhead detonation can damage the Scud warhead, raising thepossibility that the Patriot's fuze will sense the Scud too late to have any chance of hit-ting the Scud warhead.

Patriot missiles attempt to intercept missiles such as Scuds by flying the Scud'strajectory in reverse, so that the two missiles approach each other head-on, and this isthe intercept geometry we will discuss here. The large closing speed between the twomissiles (typically about 3.0-3.5 km sec-1)2 will result in the Patriot's fragment patternbeing peaked forward into a conical volume in the Scud's rest frame. Figure C-l illus-

j trates the intercept geometry.I The cone angle shown in figure C-1 depends on the speed of the Patriot's warhead! fragments relative to the closing speed between the two missiles. The high-explosive

in the Patriot warhead is Octol, and the ratio of the warhead's high explosive chargemass (C) to its fragment mass (M) is about C/M = 1.1.3 The fragment speed can then beestimated from the following equation:4

1[ C ]2

v= G ~ (C-l)1 + 0.5M

where G is the Gurney constant, which is 2,965 m sec-1 for Octol.5 The resulting frag-ment speed is 2.5 km sec-1. Since the form of the equation used here is for an infinitelylong cylinder, the actual fragment speed will be somewhat lower when end effects aretaken into account. The half-angle of the conical volume into which the fragments areconfined (in the Scud's rest frame) is then given by (X = sin-1 (fragment speed/closingspeed) = sin-1 (2.5/3.5) = 46°.6 Thus in order for an intercept attempt to succeed indestroying a Scud's warhead, for the conditions assumed here, the Patriot warheadmust detonate in a conical volume with a half-angle of 46° ahead of the Scud warhead.

The PAC-2 Patriot missile used during the Gulf War has one or more fixed strip-line fuzes along the side of the missile.7 This pulse-doppler fuze produces two beams: abroad side-looking beam for use in aircraft intercepts, and a narrower forward-lookingbeam for use in missile intercepts.8 The narrower anti-missile beam is oriented about40-55° off the nose of the Patriot.9 If the Patriot is close enough to the target missile,the fuze will detect the target as it sweeps through the narrow beam, and the Patriotwarhead will then be detonated. The Patriot fuzing geometry is illustrated in figure C-2.

Figure C-2 shows that a comparison of the cone angle into which fragments fromthe Patriot warhead will be sprayed and the cone angle formed by the radar beam ofthe Patriot fuze indicates that only near perfect and instantaneous operation of thePatriot fuze will result in any chance that fragments from the Patriot warhead will hiteven the back of the Scud warhead. For example, if the missiles are on antiparalleltrajectories but their centerlines are offset from each other by five meters, the nosetipof the Scud will first intersect the leading edge of the fuze beam when the Scud is 4.2meters ahead of the Patriot.10 (This calculation assumes that the leading edge of thefuze beam is 40° off the Patriot's nose and the fuze is located at the Patriot warheadlocation, 1.8 meters behind the nosetip of the Patriot.) Approximately one millisecondlater, with the Scud now 0.5 meters ahead of the Patriot, the Patriot warhead passes

Video Evidence on the Effectiveness of Patriot during the 1991 Gulf War 59~ r-

4m

4m

j

Warhead fra!Jmentation Vector transformation ofpattern in Patriot rest frame fragmentation pattern to

Scud rest frameFigure C-1: The geometry and timing of a head-on intercept attempt. This figure assumes aPatriot warhead fragment speed of 2.5 km sec-l and a Patriot-Scud closing speed of 3.5 kmsec-l. In the Scud's rest frame, the large closing velocity between the two missiles causes thefragment distribution to be peaked forward into a cone with a half angle of about 46°. Thus ifa fragment is to strike the Scud warhead, the Patriot warhead must detonate in a conical vol-ume ahead of the Scud warhead.

..60

Lewis and Postal

~

:ii"""::,::)! ':.:i:: '

:,:::~:...'"~I [11~~mH ~

',: :'"

Fuze beam pattern ""':"""in Patriot missile "' :""""',

""""""""""" .~"", " 1 msec = 3.5 m

"...,..'

Figure C-2: Fuzing geometry for an intercept attempt on a Scud, The closing speed isassumed to be 3.5 km sec-l; the 3.5 meter distance the missiles move relative to each other inone millisecond is shown by the arrow beneath the Scud. In order for a Patriot warhead frag-ment to hit the Scud warhead, the Patriot warhead must explode within the conical volumedefined by the solid lines. The fuze beam, oriented 40-550 off the nose of the Patriot, is shownby the dashed lines. (The fuze is assumed to be located at the same position as the Patriotwarhead,) The Patriot fuze pattern will begin to intersect the nose of the Scud only very shortlybefore the last opportunity to destroy the Scud's warhead,

out of the conical volume in which it is possible for a Patriot fragment to strike theScud warhead. If the Scud warhead is to be destroyed the Patriot must be detonatedwithin this 3.7 meter interval, which corresponds to about one millisecond of time. Ifthe offset distance between the missiles' centerlines is 20 meters instead of five, thenthe Scud nosetip first intersects the fuze beam when the Scud is 22 meters ahead of thePatriot. About two milliseconds later, with the Scud now 15 meters ahead of thePatriot, the Patriot warhead passes out of the conical volume in which a warhead killis possible.

Thus if Patriot is to kill a Scud warhead, the Patriot warhead must not explode at


the point of closest approach to the Scud warhead; it must explode ahead of the Scud.The distance ahead it must explode will be different for each different centerline offsetdistance. Moreover, a detonation behind the Scud warhead will never damage it.

The analysis here indicates that, with the assumptions used here, it is possible forthe Patriot's fuze to detonate the Patriot warhead so that fragments will strike theScud warhead. However, the margin of error is extremely slim-roughly 3.5 meters orone millisecond for a five meter offset.

However, small departures from the near-ideal parameter values assumed in ourmodel can result in conditions where the fuze cannot act quickly enough to destroy theScud warhead before it has passed by the Patriot. Two such possible non-ideal engi-neering limits on the fuze are:

.As discussed above, our estimate of the Patriot fragment velocity is probably toohigh. If the actual velocity is ten percent lower (2.25 km sec-l), then the half-angleof the conical volume in which a warhead kill is possible falls to 40O-the same asthe fuze beam leading edge-and the margin of error is cut to about 0.6 millisec-onds (independent of the offset distance).

.The pointed nose of the Scud should result in only a very small initial reflected sig-nal for the Patriot radar fuze to detect. As a result, it is unlikely that the Patriotfuze would detect the leading edge of the Scud missile. If the fuze triggers thewarhead detonation when the Scud's nosetip reaches the middle of the fuze beam(here assumed to be at 47.5°) instead of the leading edge, the margin of error willbe only about 0.5 milliseconds for a five meter offset. Combined with the effect often percent slower fragments, it would not be possible to hit the Scud warhead ifthe offset distance was eight meters or larger. If the fuze triggers only when thefull body diameter of the Scud first reaches the center of the beam pattern, it willbe too late for any fragments to strike the Scud warhead at any offset distance.

The analysis presented here, based on the best publicly available data, cannotfirmly establish over what range of closing speeds and angles Patriot would have beenable to fuze in time to have a chance of destroying a Scud's warhead. Such an analysiswould require much detailed information about the Patriot system that is not publiclyavailable, including data on the fragmentation pattern of the Patriot warhead, the pre-cise speed of fragments produced by the detonating warhead, and details of the beamgeometry of the Patriot fuzing system. In addition, the closing speed between the Scudand the Patriot, and the geometry of the Patriot-Scud encounter, will vary with eachintercept attempt, and these factors will strongly affect the possible outcomes of anintercept attempt.

Nevertheless, the analysis presented here gives a good qualitative picture of whatthe Patriot fuze had to accomplish to achieve warhead kills against Gulf War Scuds. Italso clearly indicates that accounts, such as those cited at the beginning of this appen-dix, that indicate that there was a Patriot fuze problem are entirely plausible. Thesource of such a fuzing problem would be the inability of the Patriot's fuze to sense thefront end of an Iraqi Scud before it has passed beyond the point where the Patriot'sfragmentation warhead can destroy it. The Gulf War video data we have studied,which show that in at least two engagements Patriot apparently hit a Scud withoutdestroying the Scud's warhead, is completely consistent with the possibility that thePatriot's fuze was incapable of performing adequately against the Iraqi Scud missile.

62 Lewis and Postol

An understanding of Patriot's fuzing performance is central to operating thePatriot system in the most effective way. For example, if Patriot's fuze actually wasincapable of detonating the Patriot warheads in time at closing speeds of 3.5 kIn sec-l,it might have been more effective to try to make intercepts at lower altitudes, whereclosing speeds would be smaller. Given that the performance of the Patriot fuze wasrecognized to be a potentially serious problem, one would expect it to be the subject ofdetailed analyses by the US Army both during and after the Gulf War. Instead, how-ever, according to the GAO, the Army has not used any of the (albeit limited) Gulf Wardata to analyze how Patriot's fuzing system performed:

Computer-generated data also do not provide information on whether thePatriot's fuze reacted quickly enough to destroy the Scud. A Project Officeengineer told us the closing velocity, or the speed at which the Patriot andScud approach one another, helps determine whether the Patriot's fuze hadtime to arm and detonate before the Scud passed the intercept point. He saidthis information could be determined from recorded data. However, theproject officials did not develop the information because they did not believe itwould benefit the assessment process. The additional data would not haveshown that the Patriot detonated sufficiently close to the Scud to have a highprobability of killing it.II

NOTES AND REFERENCES1. Defense News reported that according to an Israeli source, "...the Patriot fuze wasoften found to be ineffective against tactical ballistic missiles whose closing speedswere more than 2,700 meters per second. According to the Israeli source, closingspeeds in the Persian Gulf War were typically 3,600 meters per second." This reportedproblem with the Patriot fuze was also discussed in sworn testimony before the HouseGovernment Operations Committee on 7 April 1992 by Reuven Pedatzur who statedthat: "Further analysis [by a team of Israeli Air Force analysts] turned up a problemrelated to the Patriot's fuze. It was found to be unsuited to the high closing speedswhich developed between the aI-Hussein [Scud] and the Patriot."

More recently, shortcomings in Patriot's fuzing performance during the Gulf Warwere strongly implied in statements reported in Defense News about one of the mostimportant and publicized aspects of the proposed Patriot PAC-3 upgrade, which is theuse of a new multimode seeker as the Patriot warhead fuze: "[Alan] Sherer [actingdirector for theater missile defense programs at the US Army Missile Defense Com-mand] added that improvements to the Patriot's ability to aim and fuze its warheadwill provide a measure of effectiveness lacking during the gulf war. 'We want the mis-sile to be able to hit what we call the sweet spot, and that spot will be located in differ-ent places in different missiles. So knowing when to ignite the blast will be key toimproved lethality,' Sherer said."

See Barbara Opall, "Patriot Debate Resumes: Israeli Study Revives Clash on Mis-sile Performance," Defense Ne{Vs, 18 November 1991, pp. 3, 37. House GovernmentOperations Committee, Performance of the Patriot Missile, p. 129, and George Leopoldand Barbara Opall, "US Army Starts Tests on Upgraded Patriot," Defense News, 8-14February 1993, pp. 4, 52.

2. At the 7 April 1992 Congressional Hearing on the Patriot, Major General Jay M.Garner stated that: "We have a closing velocity somewhere around 9,000 miles perhour [4.0 km sec-I]." US House Government Operations Committee, Performance of

~J"


the Patriot Missile, p. 218.

3. The mass ratio is estimated from a drawing of the PAC-2 warhead in an advertis-ing brochure produced by the warhead's manufacturer. ("Patriot Antitactical MissileWarhead," Chamberlain Manufacturing Corporation, Waterloo, Iowa.) Estimates madefrom the drawing indicate that the volume ratio of high-explosives to fragments isabout 4.7. Since the fragments are made of steel, which has a density of about 7.8 gmcm-3, and the Octol high-explosive has a density of about 1.81 gm cm-3, the volumeratio indicates that the Patriot warhead explosive to fragments mass ratio is about 1.1.

4. Gilbert F. Kinney and Kenneth J. Graham, Explosive Shocks in Ail; 2nd ed. (NewYork: Springer-Verlag, 1985), p. 27; Gordon E. Jones, James E. Kennedy, and Larry D.Bertholf, "Ballistic Calculations of R.W. Gurney," American Journal of Physics, April,1980, pp. 264-269.

5. Kinney and Graham, Explosive Shocks in Air, p. 237.

6. For the geometry analyzed here, and assuming that fragments are sprayed withequal velocities in all directions, the largest half angle is given by fragments that arethrown back (in the Patriot's rest frame) at an angle of 1350 from the direction of thePatriot's motion. It is unclear if the Patriot warhead is designed to throw fragmentsback at such large angles; if it is not, then the half angle will be smaller.

7. David Hughes, "Successful Test of Advanced Patriot Confirms Value of Multi-Mode Seeker," Aviation Week & Space Technology, 27 April 1992, pp. 22-23.

8. Mark Hewish, "War-Winning Technologies: Patriot Shows its Mettle," Interna-tional Defense Review, May 1991, p. 457.

9. Hughes, "Successful Test," p. 23.

10. That is, the component of the distance between the two missiles' nosetips that isparallel to the missiles' motion is equal to 4.2 meters.

11. US General Accounting Office, "Data Does Not Exist," p. 7.

video evidence on the effectiveness of patriot...

Documents